net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static void addrconf_bonding_change(struct net_device *dev,
 141                                     unsigned long event);
 142 static int addrconf_ifdown(struct net_device *dev, int how);
 143
 144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 145 static void addrconf_dad_timer(unsigned long data);
 146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 147 static void addrconf_dad_run(struct inet6_dev *idev);
 148 static void addrconf_rs_timer(unsigned long data);
 149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 151
 152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 153                                 struct prefix_info *pinfo);
 154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 155                               struct net_device *dev);
 156
 157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 158
 159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 160         .forwarding             = 0,
 161         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 162         .mtu6                   = IPV6_MIN_MTU,
 163         .accept_ra              = 1,
 164         .accept_redirects       = 1,
 165         .autoconf               = 1,
 166         .force_mld_version      = 0,
 167         .dad_transmits          = 1,
 168         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 169         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 170         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 171 #ifdef CONFIG_IPV6_PRIVACY
 172         .use_tempaddr           = 0,
 173         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 174         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 175         .regen_max_retry        = REGEN_MAX_RETRY,
 176         .max_desync_factor      = MAX_DESYNC_FACTOR,
 177 #endif
 178         .max_addresses          = IPV6_MAX_ADDRESSES,
 179         .accept_ra_defrtr       = 1,
 180         .accept_ra_pinfo        = 1,
 181 #ifdef CONFIG_IPV6_ROUTER_PREF
 182         .accept_ra_rtr_pref     = 1,
 183         .rtr_probe_interval     = 60 * HZ,
 184 #ifdef CONFIG_IPV6_ROUTE_INFO
 185         .accept_ra_rt_info_max_plen = 0,
 186 #endif
 187 #endif
 188         .proxy_ndp              = 0,
 189         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 190         .disable_ipv6           = 0,
 191         .accept_dad             = 1,
 192 };
 193
 194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 195         .forwarding             = 0,
 196         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 197         .mtu6                   = IPV6_MIN_MTU,
 198         .accept_ra              = 1,
 199         .accept_redirects       = 1,
 200         .autoconf               = 1,
 201         .dad_transmits          = 1,
 202         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 203         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205 #ifdef CONFIG_IPV6_PRIVACY
 206         .use_tempaddr           = 0,
 207         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 208         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 209         .regen_max_retry        = REGEN_MAX_RETRY,
 210         .max_desync_factor      = MAX_DESYNC_FACTOR,
 211 #endif
 212         .max_addresses          = IPV6_MAX_ADDRESSES,
 213         .accept_ra_defrtr       = 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_max_plen = 0,
 220 #endif
 221 #endif
 222         .proxy_ndp              = 0,
 223         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 224         .disable_ipv6           = 0,
 225         .accept_dad             = 1,
 226 };
 227
 228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 233
 234 /* Check if a valid qdisc is available */
 235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 236 {
 237         return !qdisc_tx_is_noop(dev);
 238 }
 239
 240 /* Check if a route is valid prefix route */
 241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 242 {
 243         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 244 }
 245
 246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 247 {
 248         if (del_timer(&ifp->timer))
 249                 __in6_ifa_put(ifp);
 250 }
 251
 252 enum addrconf_timer_t
 253 {
 254         AC_NONE,
 255         AC_DAD,
 256         AC_RS,
 257 };
 258
 259 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 260                                enum addrconf_timer_t what,
 261                                unsigned long when)
 262 {
 263         if (!del_timer(&ifp->timer))
 264                 in6_ifa_hold(ifp);
 265
 266         switch (what) {
 267         case AC_DAD:
 268                 ifp->timer.function = addrconf_dad_timer;
 269                 break;
 270         case AC_RS:
 271                 ifp->timer.function = addrconf_rs_timer;
 272                 break;
 273         default:;
 274         }
 275         ifp->timer.expires = jiffies + when;
 276         add_timer(&ifp->timer);
 277 }
 278
 279 static int snmp6_alloc_dev(struct inet6_dev *idev)
 280 {
 281         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 282                           sizeof(struct ipstats_mib)) < 0)
 283                 goto err_ip;
 284         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 285                           sizeof(struct icmpv6_mib)) < 0)
 286                 goto err_icmp;
 287         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 288                           sizeof(struct icmpv6msg_mib)) < 0)
 289                 goto err_icmpmsg;
 290
 291         return 0;
 292
 293 err_icmpmsg:
 294         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 295 err_icmp:
 296         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 297 err_ip:
 298         return -ENOMEM;
 299 }
 300
 301 static void snmp6_free_dev(struct inet6_dev *idev)
 302 {
 303         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 306 }
 307
 308 /* Nobody refers to this device, we may destroy it. */
 309
 310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 311 {
 312         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 313         kfree(idev);
 314 }
 315
 316 void in6_dev_finish_destroy(struct inet6_dev *idev)
 317 {
 318         struct net_device *dev = idev->dev;
 319
 320         WARN_ON(idev->addr_list != NULL);
 321         WARN_ON(idev->mc_list != NULL);
 322
 323 #ifdef NET_REFCNT_DEBUG
 324         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 325 #endif
 326         dev_put(dev);
 327         if (!idev->dead) {
 328                 printk("Freeing alive inet6 device %p\n", idev);
 329                 return;
 330         }
 331         snmp6_free_dev(idev);
 332         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 333 }
 334
 335 EXPORT_SYMBOL(in6_dev_finish_destroy);
 336
 337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 338 {
 339         struct inet6_dev *ndev;
 340
 341         ASSERT_RTNL();
 342
 343         if (dev->mtu < IPV6_MIN_MTU)
 344                 return NULL;
 345
 346         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 347
 348         if (ndev == NULL)
 349                 return NULL;
 350
 351         rwlock_init(&ndev->lock);
 352         ndev->dev = dev;
 353         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 354         ndev->cnf.mtu6 = dev->mtu;
 355         ndev->cnf.sysctl = NULL;
 356         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 357         if (ndev->nd_parms == NULL) {
 358                 kfree(ndev);
 359                 return NULL;
 360         }
 361         if (ndev->cnf.forwarding)
 362                 dev_disable_lro(dev);
 363         /* We refer to the device */
 364         dev_hold(dev);
 365
 366         if (snmp6_alloc_dev(ndev) < 0) {
 367                 ADBG((KERN_WARNING
 368                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 369                         __func__, dev->name));
 370                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 371                 ndev->dead = 1;
 372                 in6_dev_finish_destroy(ndev);
 373                 return NULL;
 374         }
 375
 376         if (snmp6_register_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 ndev->dead = 1;
 382                 in6_dev_finish_destroy(ndev);
 383                 return NULL;
 384         }
 385
 386         /* One reference from device.  We must do this before
 387          * we invoke __ipv6_regen_rndid().
 388          */
 389         in6_dev_hold(ndev);
 390
 391         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 392                 ndev->cnf.accept_dad = -1;
 393
 394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 395         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 396                 printk(KERN_INFO
 397                        "%s: Disabled Multicast RS\n",
 398                        dev->name);
 399                 ndev->cnf.rtr_solicits = 0;
 400         }
 401 #endif
 402
 403 #ifdef CONFIG_IPV6_PRIVACY
 404         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 405         if ((dev->flags&IFF_LOOPBACK) ||
 406             dev->type == ARPHRD_TUNNEL ||
 407             dev->type == ARPHRD_TUNNEL6 ||
 408             dev->type == ARPHRD_SIT ||
 409             dev->type == ARPHRD_NONE) {
 410                 printk(KERN_INFO
 411                        "%s: Disabled Privacy Extensions\n",
 412                        dev->name);
 413                 ndev->cnf.use_tempaddr = -1;
 414         } else {
 415                 in6_dev_hold(ndev);
 416                 ipv6_regen_rndid((unsigned long) ndev);
 417         }
 418 #endif
 419
 420         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 421                 ndev->if_flags |= IF_READY;
 422
 423         ipv6_mc_init_dev(ndev);
 424         ndev->tstamp = jiffies;
 425         addrconf_sysctl_register(ndev);
 426         /* protected by rtnl_lock */
 427         rcu_assign_pointer(dev->ip6_ptr, ndev);
 428
 429         /* Join all-node multicast group */
 430         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 431
 432         return ndev;
 433 }
 434
 435 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 436 {
 437         struct inet6_dev *idev;
 438
 439         ASSERT_RTNL();
 440
 441         if ((idev = __in6_dev_get(dev)) == NULL) {
 442                 if ((idev = ipv6_add_dev(dev)) == NULL)
 443                         return NULL;
 444         }
 445
 446         if (dev->flags&IFF_UP)
 447                 ipv6_mc_up(idev);
 448         return idev;
 449 }
 450
 451 #ifdef CONFIG_SYSCTL
 452 static void dev_forward_change(struct inet6_dev *idev)
 453 {
 454         struct net_device *dev;
 455         struct inet6_ifaddr *ifa;
 456
 457         if (!idev)
 458                 return;
 459         dev = idev->dev;
 460         if (idev->cnf.forwarding)
 461                 dev_disable_lro(dev);
 462         if (dev && (dev->flags & IFF_MULTICAST)) {
 463                 if (idev->cnf.forwarding)
 464                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 465                 else
 466                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 467         }
 468         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 469                 if (ifa->flags&IFA_F_TENTATIVE)
 470                         continue;
 471                 if (idev->cnf.forwarding)
 472                         addrconf_join_anycast(ifa);
 473                 else
 474                         addrconf_leave_anycast(ifa);
 475         }
 476 }
 477
 478
 479 static void addrconf_forward_change(struct net *net, __s32 newf)
 480 {
 481         struct net_device *dev;
 482         struct inet6_dev *idev;
 483
 484         rcu_read_lock();
 485         for_each_netdev_rcu(net, dev) {
 486                 idev = __in6_dev_get(dev);
 487                 if (idev) {
 488                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 489                         idev->cnf.forwarding = newf;
 490                         if (changed)
 491                                 dev_forward_change(idev);
 492                 }
 493         }
 494         rcu_read_unlock();
 495 }
 496
 497 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 498 {
 499         struct net *net;
 500
 501         net = (struct net *)table->extra2;
 502         if (p == &net->ipv6.devconf_dflt->forwarding)
 503                 return 0;
 504
 505         if (!rtnl_trylock()) {
 506                 /* Restore the original values before restarting */
 507                 *p = old;
 508                 return restart_syscall();
 509         }
 510
 511         if (p == &net->ipv6.devconf_all->forwarding) {
 512                 __s32 newf = net->ipv6.devconf_all->forwarding;
 513                 net->ipv6.devconf_dflt->forwarding = newf;
 514                 addrconf_forward_change(net, newf);
 515         } else if ((!*p) ^ (!old))
 516                 dev_forward_change((struct inet6_dev *)table->extra1);
 517         rtnl_unlock();
 518
 519         if (*p)
 520                 rt6_purge_dflt_routers(net);
 521         return 1;
 522 }
 523 #endif
 524
 525 /* Nobody refers to this ifaddr, destroy it */
 526
 527 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 528 {
 529         WARN_ON(ifp->if_next != NULL);
 530         WARN_ON(ifp->lst_next != NULL);
 531
 532 #ifdef NET_REFCNT_DEBUG
 533         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 534 #endif
 535
 536         in6_dev_put(ifp->idev);
 537
 538         if (del_timer(&ifp->timer))
 539                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 540
 541         if (!ifp->dead) {
 542                 printk("Freeing alive inet6 address %p\n", ifp);
 543                 return;
 544         }
 545         dst_release(&ifp->rt->u.dst);
 546
 547         kfree(ifp);
 548 }
 549
 550 static void
 551 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 552 {
 553         struct inet6_ifaddr *ifa, **ifap;
 554         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 555
 556         /*
 557          * Each device address list is sorted in order of scope -
 558          * global before linklocal.
 559          */
 560         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 561              ifap = &ifa->if_next) {
 562                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 563                         break;
 564         }
 565
 566         ifp->if_next = *ifap;
 567         *ifap = ifp;
 568 }
 569
 570 /*
 571  *      Hash function taken from net_alias.c
 572  */
 573 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 574 {
 575         __u32 word;
 576
 577         /*
 578          * We perform the hash function over the last 64 bits of the address
 579          * This will include the IEEE address token on links that support it.
 580          */
 581
 582         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 583         word ^= (word >> 16);
 584         word ^= (word >> 8);
 585
 586         return ((word ^ (word >> 4)) & 0x0f);
 587 }
 588
 589 /* On success it returns ifp with increased reference count */
 590
 591 static struct inet6_ifaddr *
 592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 593               int scope, u32 flags)
 594 {
 595         struct inet6_ifaddr *ifa = NULL;
 596         struct rt6_info *rt;
 597         int hash;
 598         int err = 0;
 599         int addr_type = ipv6_addr_type(addr);
 600
 601         if (addr_type == IPV6_ADDR_ANY ||
 602             addr_type & IPV6_ADDR_MULTICAST ||
 603             (!(idev->dev->flags & IFF_LOOPBACK) &&
 604              addr_type & IPV6_ADDR_LOOPBACK))
 605                 return ERR_PTR(-EADDRNOTAVAIL);
 606
 607         rcu_read_lock_bh();
 608         if (idev->dead) {
 609                 err = -ENODEV;                  /*XXX*/
 610                 goto out2;
 611         }
 612
 613         if (idev->cnf.disable_ipv6) {
 614                 err = -EACCES;
 615                 goto out2;
 616         }
 617
 618         write_lock(&addrconf_hash_lock);
 619
 620         /* Ignore adding duplicate addresses on an interface */
 621         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 622                 ADBG(("ipv6_add_addr: already assigned\n"));
 623                 err = -EEXIST;
 624                 goto out;
 625         }
 626
 627         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 628
 629         if (ifa == NULL) {
 630                 ADBG(("ipv6_add_addr: malloc failed\n"));
 631                 err = -ENOBUFS;
 632                 goto out;
 633         }
 634
 635         rt = addrconf_dst_alloc(idev, addr, 0);
 636         if (IS_ERR(rt)) {
 637                 err = PTR_ERR(rt);
 638                 goto out;
 639         }
 640
 641         ipv6_addr_copy(&ifa->addr, addr);
 642
 643         spin_lock_init(&ifa->lock);
 644         init_timer(&ifa->timer);
 645         ifa->timer.data = (unsigned long) ifa;
 646         ifa->scope = scope;
 647         ifa->prefix_len = pfxlen;
 648         ifa->flags = flags | IFA_F_TENTATIVE;
 649         ifa->cstamp = ifa->tstamp = jiffies;
 650
 651         ifa->rt = rt;
 652
 653         /*
 654          * part one of RFC 4429, section 3.3
 655          * We should not configure an address as
 656          * optimistic if we do not yet know the link
 657          * layer address of our nexhop router
 658          */
 659
 660         if (rt->rt6i_nexthop == NULL)
 661                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 662
 663         ifa->idev = idev;
 664         in6_dev_hold(idev);
 665         /* For caller */
 666         in6_ifa_hold(ifa);
 667
 668         /* Add to big hash table */
 669         hash = ipv6_addr_hash(addr);
 670
 671         ifa->lst_next = inet6_addr_lst[hash];
 672         inet6_addr_lst[hash] = ifa;
 673         in6_ifa_hold(ifa);
 674         write_unlock(&addrconf_hash_lock);
 675
 676         write_lock(&idev->lock);
 677         /* Add to inet6_dev unicast addr list. */
 678         ipv6_link_dev_addr(idev, ifa);
 679
 680 #ifdef CONFIG_IPV6_PRIVACY
 681         if (ifa->flags&IFA_F_TEMPORARY) {
 682                 ifa->tmp_next = idev->tempaddr_list;
 683                 idev->tempaddr_list = ifa;
 684                 in6_ifa_hold(ifa);
 685         }
 686 #endif
 687
 688         in6_ifa_hold(ifa);
 689         write_unlock(&idev->lock);
 690 out2:
 691         rcu_read_unlock_bh();
 692
 693         if (likely(err == 0))
 694                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 695         else {
 696                 kfree(ifa);
 697                 ifa = ERR_PTR(err);
 698         }
 699
 700         return ifa;
 701 out:
 702         write_unlock(&addrconf_hash_lock);
 703         goto out2;
 704 }
 705
 706 /* This function wants to get referenced ifp and releases it before return */
 707
 708 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 709 {
 710         struct inet6_ifaddr *ifa, **ifap;
 711         struct inet6_dev *idev = ifp->idev;
 712         int hash;
 713         int deleted = 0, onlink = 0;
 714         unsigned long expires = jiffies;
 715
 716         hash = ipv6_addr_hash(&ifp->addr);
 717
 718         ifp->dead = 1;
 719
 720         write_lock_bh(&addrconf_hash_lock);
 721         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 722              ifap = &ifa->lst_next) {
 723                 if (ifa == ifp) {
 724                         *ifap = ifa->lst_next;
 725                         __in6_ifa_put(ifp);
 726                         ifa->lst_next = NULL;
 727                         break;
 728                 }
 729         }
 730         write_unlock_bh(&addrconf_hash_lock);
 731
 732         write_lock_bh(&idev->lock);
 733 #ifdef CONFIG_IPV6_PRIVACY
 734         if (ifp->flags&IFA_F_TEMPORARY) {
 735                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 736                      ifap = &ifa->tmp_next) {
 737                         if (ifa == ifp) {
 738                                 *ifap = ifa->tmp_next;
 739                                 if (ifp->ifpub) {
 740                                         in6_ifa_put(ifp->ifpub);
 741                                         ifp->ifpub = NULL;
 742                                 }
 743                                 __in6_ifa_put(ifp);
 744                                 ifa->tmp_next = NULL;
 745                                 break;
 746                         }
 747                 }
 748         }
 749 #endif
 750
 751         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 752                 if (ifa == ifp) {
 753                         *ifap = ifa->if_next;
 754                         __in6_ifa_put(ifp);
 755                         ifa->if_next = NULL;
 756                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 757                                 break;
 758                         deleted = 1;
 759                         continue;
 760                 } else if (ifp->flags & IFA_F_PERMANENT) {
 761                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 762                                               ifp->prefix_len)) {
 763                                 if (ifa->flags & IFA_F_PERMANENT) {
 764                                         onlink = 1;
 765                                         if (deleted)
 766                                                 break;
 767                                 } else {
 768                                         unsigned long lifetime;
 769
 770                                         if (!onlink)
 771                                                 onlink = -1;
 772
 773                                         spin_lock(&ifa->lock);
 774
 775                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 776                                         /*
 777                                          * Note: Because this address is
 778                                          * not permanent, lifetime <
 779                                          * LONG_MAX / HZ here.
 780                                          */
 781                                         if (time_before(expires,
 782                                                         ifa->tstamp + lifetime * HZ))
 783                                                 expires = ifa->tstamp + lifetime * HZ;
 784                                         spin_unlock(&ifa->lock);
 785                                 }
 786                         }
 787                 }
 788                 ifap = &ifa->if_next;
 789         }
 790         write_unlock_bh(&idev->lock);
 791
 792         addrconf_del_timer(ifp);
 793
 794         ipv6_ifa_notify(RTM_DELADDR, ifp);
 795
 796         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 797
 798         /*
 799          * Purge or update corresponding prefix
 800          *
 801          * 1) we don't purge prefix here if address was not permanent.
 802          *    prefix is managed by its own lifetime.
 803          * 2) if there're no addresses, delete prefix.
 804          * 3) if there're still other permanent address(es),
 805          *    corresponding prefix is still permanent.
 806          * 4) otherwise, update prefix lifetime to the
 807          *    longest valid lifetime among the corresponding
 808          *    addresses on the device.
 809          *    Note: subsequent RA will update lifetime.
 810          *
 811          * --yoshfuji
 812          */
 813         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 814                 struct in6_addr prefix;
 815                 struct rt6_info *rt;
 816                 struct net *net = dev_net(ifp->idev->dev);
 817                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 818                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 819
 820                 if (rt && addrconf_is_prefix_route(rt)) {
 821                         if (onlink == 0) {
 822                                 ip6_del_rt(rt);
 823                                 rt = NULL;
 824                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 825                                 rt->rt6i_expires = expires;
 826                                 rt->rt6i_flags |= RTF_EXPIRES;
 827                         }
 828                 }
 829                 dst_release(&rt->u.dst);
 830         }
 831
 832         in6_ifa_put(ifp);
 833 }
 834
 835 #ifdef CONFIG_IPV6_PRIVACY
 836 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 837 {
 838         struct inet6_dev *idev = ifp->idev;
 839         struct in6_addr addr, *tmpaddr;
 840         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 841         unsigned long regen_advance;
 842         int tmp_plen;
 843         int ret = 0;
 844         int max_addresses;
 845         u32 addr_flags;
 846
 847         write_lock(&idev->lock);
 848         if (ift) {
 849                 spin_lock_bh(&ift->lock);
 850                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 851                 spin_unlock_bh(&ift->lock);
 852                 tmpaddr = &addr;
 853         } else {
 854                 tmpaddr = NULL;
 855         }
 856 retry:
 857         in6_dev_hold(idev);
 858         if (idev->cnf.use_tempaddr <= 0) {
 859                 write_unlock(&idev->lock);
 860                 printk(KERN_INFO
 861                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 862                 in6_dev_put(idev);
 863                 ret = -1;
 864                 goto out;
 865         }
 866         spin_lock_bh(&ifp->lock);
 867         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 868                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 869                 spin_unlock_bh(&ifp->lock);
 870                 write_unlock(&idev->lock);
 871                 printk(KERN_WARNING
 872                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 873                 in6_dev_put(idev);
 874                 ret = -1;
 875                 goto out;
 876         }
 877         in6_ifa_hold(ifp);
 878         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 879         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 880                 spin_unlock_bh(&ifp->lock);
 881                 write_unlock(&idev->lock);
 882                 printk(KERN_WARNING
 883                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 884                 in6_ifa_put(ifp);
 885                 in6_dev_put(idev);
 886                 ret = -1;
 887                 goto out;
 888         }
 889         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 890         tmp_valid_lft = min_t(__u32,
 891                               ifp->valid_lft,
 892                               idev->cnf.temp_valid_lft);
 893         tmp_prefered_lft = min_t(__u32,
 894                                  ifp->prefered_lft,
 895                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 896         tmp_plen = ifp->prefix_len;
 897         max_addresses = idev->cnf.max_addresses;
 898         tmp_cstamp = ifp->cstamp;
 899         tmp_tstamp = ifp->tstamp;
 900         spin_unlock_bh(&ifp->lock);
 901
 902         regen_advance = idev->cnf.regen_max_retry *
 903                         idev->cnf.dad_transmits *
 904                         idev->nd_parms->retrans_time / HZ;
 905         write_unlock(&idev->lock);
 906
 907         /* A temporary address is created only if this calculated Preferred
 908          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 909          * an implementation must not create a temporary address with a zero
 910          * Preferred Lifetime.
 911          */
 912         if (tmp_prefered_lft <= regen_advance) {
 913                 in6_ifa_put(ifp);
 914                 in6_dev_put(idev);
 915                 ret = -1;
 916                 goto out;
 917         }
 918
 919         addr_flags = IFA_F_TEMPORARY;
 920         /* set in addrconf_prefix_rcv() */
 921         if (ifp->flags & IFA_F_OPTIMISTIC)
 922                 addr_flags |= IFA_F_OPTIMISTIC;
 923
 924         ift = !max_addresses ||
 925               ipv6_count_addresses(idev) < max_addresses ?
 926                 ipv6_add_addr(idev, &addr, tmp_plen,
 927                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 928                               addr_flags) : NULL;
 929         if (!ift || IS_ERR(ift)) {
 930                 in6_ifa_put(ifp);
 931                 in6_dev_put(idev);
 932                 printk(KERN_INFO
 933                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 934                 tmpaddr = &addr;
 935                 write_lock(&idev->lock);
 936                 goto retry;
 937         }
 938
 939         spin_lock_bh(&ift->lock);
 940         ift->ifpub = ifp;
 941         ift->valid_lft = tmp_valid_lft;
 942         ift->prefered_lft = tmp_prefered_lft;
 943         ift->cstamp = tmp_cstamp;
 944         ift->tstamp = tmp_tstamp;
 945         spin_unlock_bh(&ift->lock);
 946
 947         addrconf_dad_start(ift, 0);
 948         in6_ifa_put(ift);
 949         in6_dev_put(idev);
 950 out:
 951         return ret;
 952 }
 953 #endif
 954
 955 /*
 956  *      Choose an appropriate source address (RFC3484)
 957  */
 958 enum {
 959         IPV6_SADDR_RULE_INIT = 0,
 960         IPV6_SADDR_RULE_LOCAL,
 961         IPV6_SADDR_RULE_SCOPE,
 962         IPV6_SADDR_RULE_PREFERRED,
 963 #ifdef CONFIG_IPV6_MIP6
 964         IPV6_SADDR_RULE_HOA,
 965 #endif
 966         IPV6_SADDR_RULE_OIF,
 967         IPV6_SADDR_RULE_LABEL,
 968 #ifdef CONFIG_IPV6_PRIVACY
 969         IPV6_SADDR_RULE_PRIVACY,
 970 #endif
 971         IPV6_SADDR_RULE_ORCHID,
 972         IPV6_SADDR_RULE_PREFIX,
 973         IPV6_SADDR_RULE_MAX
 974 };
 975
 976 struct ipv6_saddr_score {
 977         int                     rule;
 978         int                     addr_type;
 979         struct inet6_ifaddr     *ifa;
 980         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 981         int                     scopedist;
 982         int                     matchlen;
 983 };
 984
 985 struct ipv6_saddr_dst {
 986         const struct in6_addr *addr;
 987         int ifindex;
 988         int scope;
 989         int label;
 990         unsigned int prefs;
 991 };
 992
 993 static inline int ipv6_saddr_preferred(int type)
 994 {
 995         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 996                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 997                 return 1;
 998         return 0;
 999 }
1000
1001 static int ipv6_get_saddr_eval(struct net *net,
1002                                struct ipv6_saddr_score *score,
1003                                struct ipv6_saddr_dst *dst,
1004                                int i)
1005 {
1006         int ret;
1007
1008         if (i <= score->rule) {
1009                 switch (i) {
1010                 case IPV6_SADDR_RULE_SCOPE:
1011                         ret = score->scopedist;
1012                         break;
1013                 case IPV6_SADDR_RULE_PREFIX:
1014                         ret = score->matchlen;
1015                         break;
1016                 default:
1017                         ret = !!test_bit(i, score->scorebits);
1018                 }
1019                 goto out;
1020         }
1021
1022         switch (i) {
1023         case IPV6_SADDR_RULE_INIT:
1024                 /* Rule 0: remember if hiscore is not ready yet */
1025                 ret = !!score->ifa;
1026                 break;
1027         case IPV6_SADDR_RULE_LOCAL:
1028                 /* Rule 1: Prefer same address */
1029                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1030                 break;
1031         case IPV6_SADDR_RULE_SCOPE:
1032                 /* Rule 2: Prefer appropriate scope
1033                  *
1034                  *      ret
1035                  *       ^
1036                  *    -1 |  d 15
1037                  *    ---+--+-+---> scope
1038                  *       |
1039                  *       |             d is scope of the destination.
1040                  *  B-d  |  \
1041                  *       |   \      <- smaller scope is better if
1042                  *  B-15 |    \        if scope is enough for destinaion.
1043                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1044                  * d-C-1 | /
1045                  *       |/         <- greater is better
1046                  *   -C  /             if scope is not enough for destination.
1047                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1048                  *
1049                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1050                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1051                  * Assume B = 0 and we get C > 29.
1052                  */
1053                 ret = __ipv6_addr_src_scope(score->addr_type);
1054                 if (ret >= dst->scope)
1055                         ret = -ret;
1056                 else
1057                         ret -= 128;     /* 30 is enough */
1058                 score->scopedist = ret;
1059                 break;
1060         case IPV6_SADDR_RULE_PREFERRED:
1061                 /* Rule 3: Avoid deprecated and optimistic addresses */
1062                 ret = ipv6_saddr_preferred(score->addr_type) ||
1063                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1064                 break;
1065 #ifdef CONFIG_IPV6_MIP6
1066         case IPV6_SADDR_RULE_HOA:
1067             {
1068                 /* Rule 4: Prefer home address */
1069                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1070                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1071                 break;
1072             }
1073 #endif
1074         case IPV6_SADDR_RULE_OIF:
1075                 /* Rule 5: Prefer outgoing interface */
1076                 ret = (!dst->ifindex ||
1077                        dst->ifindex == score->ifa->idev->dev->ifindex);
1078                 break;
1079         case IPV6_SADDR_RULE_LABEL:
1080                 /* Rule 6: Prefer matching label */
1081                 ret = ipv6_addr_label(net,
1082                                       &score->ifa->addr, score->addr_type,
1083                                       score->ifa->idev->dev->ifindex) == dst->label;
1084                 break;
1085 #ifdef CONFIG_IPV6_PRIVACY
1086         case IPV6_SADDR_RULE_PRIVACY:
1087             {
1088                 /* Rule 7: Prefer public address
1089                  * Note: prefer temprary address if use_tempaddr >= 2
1090                  */
1091                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1092                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1093                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1094                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1095                 break;
1096             }
1097 #endif
1098         case IPV6_SADDR_RULE_ORCHID:
1099                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1100                  *          non-ORCHID vs non-ORCHID
1101                  */
1102                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1103                         ipv6_addr_orchid(dst->addr));
1104                 break;
1105         case IPV6_SADDR_RULE_PREFIX:
1106                 /* Rule 8: Use longest matching prefix */
1107                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1108                                                        dst->addr);
1109                 break;
1110         default:
1111                 ret = 0;
1112         }
1113
1114         if (ret)
1115                 __set_bit(i, score->scorebits);
1116         score->rule = i;
1117 out:
1118         return ret;
1119 }
1120
1121 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1122                        const struct in6_addr *daddr, unsigned int prefs,
1123                        struct in6_addr *saddr)
1124 {
1125         struct ipv6_saddr_score scores[2],
1126                                 *score = &scores[0], *hiscore = &scores[1];
1127         struct ipv6_saddr_dst dst;
1128         struct net_device *dev;
1129         int dst_type;
1130
1131         dst_type = __ipv6_addr_type(daddr);
1132         dst.addr = daddr;
1133         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1134         dst.scope = __ipv6_addr_src_scope(dst_type);
1135         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1136         dst.prefs = prefs;
1137
1138         hiscore->rule = -1;
1139         hiscore->ifa = NULL;
1140
1141         rcu_read_lock();
1142
1143         for_each_netdev_rcu(net, dev) {
1144                 struct inet6_dev *idev;
1145
1146                 /* Candidate Source Address (section 4)
1147                  *  - multicast and link-local destination address,
1148                  *    the set of candidate source address MUST only
1149                  *    include addresses assigned to interfaces
1150                  *    belonging to the same link as the outgoing
1151                  *    interface.
1152                  * (- For site-local destination addresses, the
1153                  *    set of candidate source addresses MUST only
1154                  *    include addresses assigned to interfaces
1155                  *    belonging to the same site as the outgoing
1156                  *    interface.)
1157                  */
1158                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160                     dst.ifindex && dev->ifindex != dst.ifindex)
1161                         continue;
1162
1163                 idev = __in6_dev_get(dev);
1164                 if (!idev)
1165                         continue;
1166
1167                 read_lock_bh(&idev->lock);
1168                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1169                         int i;
1170
1171                         /*
1172                          * - Tentative Address (RFC2462 section 5.4)
1173                          *  - A tentative address is not considered
1174                          *    "assigned to an interface" in the traditional
1175                          *    sense, unless it is also flagged as optimistic.
1176                          * - Candidate Source Address (section 4)
1177                          *  - In any case, anycast addresses, multicast
1178                          *    addresses, and the unspecified address MUST
1179                          *    NOT be included in a candidate set.
1180                          */
1181                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183                                 continue;
1184
1185                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1186
1187                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1189                                 LIMIT_NETDEBUG(KERN_DEBUG
1190                                                "ADDRCONF: unspecified / multicast address "
1191                                                "assigned as unicast address on %s",
1192                                                dev->name);
1193                                 continue;
1194                         }
1195
1196                         score->rule = -1;
1197                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1198
1199                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200                                 int minihiscore, miniscore;
1201
1202                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1204
1205                                 if (minihiscore > miniscore) {
1206                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1207                                             score->scopedist > 0) {
1208                                                 /*
1209                                                  * special case:
1210                                                  * each remaining entry
1211                                                  * has too small (not enough)
1212                                                  * scope, because ifa entries
1213                                                  * are sorted by their scope
1214                                                  * values.
1215                                                  */
1216                                                 goto try_nextdev;
1217                                         }
1218                                         break;
1219                                 } else if (minihiscore < miniscore) {
1220                                         if (hiscore->ifa)
1221                                                 in6_ifa_put(hiscore->ifa);
1222
1223                                         in6_ifa_hold(score->ifa);
1224
1225                                         swap(hiscore, score);
1226
1227                                         /* restore our iterator */
1228                                         score->ifa = hiscore->ifa;
1229
1230                                         break;
1231                                 }
1232                         }
1233                 }
1234 try_nextdev:
1235                 read_unlock_bh(&idev->lock);
1236         }
1237         rcu_read_unlock();
1238
1239         if (!hiscore->ifa)
1240                 return -EADDRNOTAVAIL;
1241
1242         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243         in6_ifa_put(hiscore->ifa);
1244         return 0;
1245 }
1246
1247 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1248
1249 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1250                     unsigned char banned_flags)
1251 {
1252         struct inet6_dev *idev;
1253         int err = -EADDRNOTAVAIL;
1254
1255         rcu_read_lock();
1256         if ((idev = __in6_dev_get(dev)) != NULL) {
1257                 struct inet6_ifaddr *ifp;
1258
1259                 read_lock_bh(&idev->lock);
1260                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1261                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1262                                 ipv6_addr_copy(addr, &ifp->addr);
1263                                 err = 0;
1264                                 break;
1265                         }
1266                 }
1267                 read_unlock_bh(&idev->lock);
1268         }
1269         rcu_read_unlock();
1270         return err;
1271 }
1272
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 {
1275         int cnt = 0;
1276         struct inet6_ifaddr *ifp;
1277
1278         read_lock_bh(&idev->lock);
1279         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1280                 cnt++;
1281         read_unlock_bh(&idev->lock);
1282         return cnt;
1283 }
1284
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286                   struct net_device *dev, int strict)
1287 {
1288         struct inet6_ifaddr * ifp;
1289         u8 hash = ipv6_addr_hash(addr);
1290
1291         read_lock_bh(&addrconf_hash_lock);
1292         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1293                 if (!net_eq(dev_net(ifp->idev->dev), net))
1294                         continue;
1295                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1296                     !(ifp->flags&IFA_F_TENTATIVE)) {
1297                         if (dev == NULL || ifp->idev->dev == dev ||
1298                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1299                                 break;
1300                 }
1301         }
1302         read_unlock_bh(&addrconf_hash_lock);
1303         return ifp != NULL;
1304 }
1305 EXPORT_SYMBOL(ipv6_chk_addr);
1306
1307 static
1308 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309                        struct net_device *dev)
1310 {
1311         struct inet6_ifaddr * ifp;
1312         u8 hash = ipv6_addr_hash(addr);
1313
1314         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1315                 if (!net_eq(dev_net(ifp->idev->dev), net))
1316                         continue;
1317                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1318                         if (dev == NULL || ifp->idev->dev == dev)
1319                                 break;
1320                 }
1321         }
1322         return ifp != NULL;
1323 }
1324
1325 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1326 {
1327         struct inet6_dev *idev;
1328         struct inet6_ifaddr *ifa;
1329         int     onlink;
1330
1331         onlink = 0;
1332         rcu_read_lock();
1333         idev = __in6_dev_get(dev);
1334         if (idev) {
1335                 read_lock_bh(&idev->lock);
1336                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1337                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1338                                                    ifa->prefix_len);
1339                         if (onlink)
1340                                 break;
1341                 }
1342                 read_unlock_bh(&idev->lock);
1343         }
1344         rcu_read_unlock();
1345         return onlink;
1346 }
1347
1348 EXPORT_SYMBOL(ipv6_chk_prefix);
1349
1350 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1351                                      struct net_device *dev, int strict)
1352 {
1353         struct inet6_ifaddr * ifp;
1354         u8 hash = ipv6_addr_hash(addr);
1355
1356         read_lock_bh(&addrconf_hash_lock);
1357         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1358                 if (!net_eq(dev_net(ifp->idev->dev), net))
1359                         continue;
1360                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1361                         if (dev == NULL || ifp->idev->dev == dev ||
1362                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1363                                 in6_ifa_hold(ifp);
1364                                 break;
1365                         }
1366                 }
1367         }
1368         read_unlock_bh(&addrconf_hash_lock);
1369
1370         return ifp;
1371 }
1372
1373 /* Gets referenced address, destroys ifaddr */
1374
1375 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1376 {
1377         if (ifp->flags&IFA_F_PERMANENT) {
1378                 spin_lock_bh(&ifp->lock);
1379                 addrconf_del_timer(ifp);
1380                 ifp->flags |= IFA_F_TENTATIVE;
1381                 if (dad_failed)
1382                         ifp->flags |= IFA_F_DADFAILED;
1383                 spin_unlock_bh(&ifp->lock);
1384                 in6_ifa_put(ifp);
1385 #ifdef CONFIG_IPV6_PRIVACY
1386         } else if (ifp->flags&IFA_F_TEMPORARY) {
1387                 struct inet6_ifaddr *ifpub;
1388                 spin_lock_bh(&ifp->lock);
1389                 ifpub = ifp->ifpub;
1390                 if (ifpub) {
1391                         in6_ifa_hold(ifpub);
1392                         spin_unlock_bh(&ifp->lock);
1393                         ipv6_create_tempaddr(ifpub, ifp);
1394                         in6_ifa_put(ifpub);
1395                 } else {
1396                         spin_unlock_bh(&ifp->lock);
1397                 }
1398                 ipv6_del_addr(ifp);
1399 #endif
1400         } else
1401                 ipv6_del_addr(ifp);
1402 }
1403
1404 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1405 {
1406         struct inet6_dev *idev = ifp->idev;
1407
1408         if (net_ratelimit())
1409                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1410                         ifp->idev->dev->name, &ifp->addr);
1411
1412         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1413                 struct in6_addr addr;
1414
1415                 addr.s6_addr32[0] = htonl(0xfe800000);
1416                 addr.s6_addr32[1] = 0;
1417
1418                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1419                     ipv6_addr_equal(&ifp->addr, &addr)) {
1420                         /* DAD failed for link-local based on MAC address */
1421                         idev->cnf.disable_ipv6 = 1;
1422
1423                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1424                                 ifp->idev->dev->name);
1425                 }
1426         }
1427
1428         addrconf_dad_stop(ifp, 1);
1429 }
1430
1431 /* Join to solicited addr multicast group. */
1432
1433 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1434 {
1435         struct in6_addr maddr;
1436
1437         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1438                 return;
1439
1440         addrconf_addr_solict_mult(addr, &maddr);
1441         ipv6_dev_mc_inc(dev, &maddr);
1442 }
1443
1444 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1445 {
1446         struct in6_addr maddr;
1447
1448         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1449                 return;
1450
1451         addrconf_addr_solict_mult(addr, &maddr);
1452         __ipv6_dev_mc_dec(idev, &maddr);
1453 }
1454
1455 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1456 {
1457         struct in6_addr addr;
1458         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1459         if (ipv6_addr_any(&addr))
1460                 return;
1461         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1462 }
1463
1464 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1465 {
1466         struct in6_addr addr;
1467         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1468         if (ipv6_addr_any(&addr))
1469                 return;
1470         __ipv6_dev_ac_dec(ifp->idev, &addr);
1471 }
1472
1473 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1474 {
1475         if (dev->addr_len != ETH_ALEN)
1476                 return -1;
1477         memcpy(eui, dev->dev_addr, 3);
1478         memcpy(eui + 5, dev->dev_addr + 3, 3);
1479
1480         /*
1481          * The zSeries OSA network cards can be shared among various
1482          * OS instances, but the OSA cards have only one MAC address.
1483          * This leads to duplicate address conflicts in conjunction
1484          * with IPv6 if more than one instance uses the same card.
1485          *
1486          * The driver for these cards can deliver a unique 16-bit
1487          * identifier for each instance sharing the same card.  It is
1488          * placed instead of 0xFFFE in the interface identifier.  The
1489          * "u" bit of the interface identifier is not inverted in this
1490          * case.  Hence the resulting interface identifier has local
1491          * scope according to RFC2373.
1492          */
1493         if (dev->dev_id) {
1494                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1495                 eui[4] = dev->dev_id & 0xFF;
1496         } else {
1497                 eui[3] = 0xFF;
1498                 eui[4] = 0xFE;
1499                 eui[0] ^= 2;
1500         }
1501         return 0;
1502 }
1503
1504 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1505 {
1506         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1507         if (dev->addr_len != ARCNET_ALEN)
1508                 return -1;
1509         memset(eui, 0, 7);
1510         eui[7] = *(u8*)dev->dev_addr;
1511         return 0;
1512 }
1513
1514 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1515 {
1516         if (dev->addr_len != INFINIBAND_ALEN)
1517                 return -1;
1518         memcpy(eui, dev->dev_addr + 12, 8);
1519         eui[0] |= 2;
1520         return 0;
1521 }
1522
1523 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1524 {
1525         if (addr == 0)
1526                 return -1;
1527         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1528                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1529                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1530                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1531                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1532                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1533         eui[1] = 0;
1534         eui[2] = 0x5E;
1535         eui[3] = 0xFE;
1536         memcpy(eui + 4, &addr, 4);
1537         return 0;
1538 }
1539 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1540
1541 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1542 {
1543         if (dev->priv_flags & IFF_ISATAP)
1544                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1545         return -1;
1546 }
1547
1548 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1549 {
1550         switch (dev->type) {
1551         case ARPHRD_ETHER:
1552         case ARPHRD_FDDI:
1553         case ARPHRD_IEEE802_TR:
1554                 return addrconf_ifid_eui48(eui, dev);
1555         case ARPHRD_ARCNET:
1556                 return addrconf_ifid_arcnet(eui, dev);
1557         case ARPHRD_INFINIBAND:
1558                 return addrconf_ifid_infiniband(eui, dev);
1559         case ARPHRD_SIT:
1560                 return addrconf_ifid_sit(eui, dev);
1561         }
1562         return -1;
1563 }
1564
1565 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1566 {
1567         int err = -1;
1568         struct inet6_ifaddr *ifp;
1569
1570         read_lock_bh(&idev->lock);
1571         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1572                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1573                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1574                         err = 0;
1575                         break;
1576                 }
1577         }
1578         read_unlock_bh(&idev->lock);
1579         return err;
1580 }
1581
1582 #ifdef CONFIG_IPV6_PRIVACY
1583 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1584 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1585 {
1586 regen:
1587         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1588         idev->rndid[0] &= ~0x02;
1589
1590         /*
1591          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1592          * check if generated address is not inappropriate
1593          *
1594          *  - Reserved subnet anycast (RFC 2526)
1595          *      11111101 11....11 1xxxxxxx
1596          *  - ISATAP (RFC4214) 6.1
1597          *      00-00-5E-FE-xx-xx-xx-xx
1598          *  - value 0
1599          *  - XXX: already assigned to an address on the device
1600          */
1601         if (idev->rndid[0] == 0xfd &&
1602             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1603             (idev->rndid[7]&0x80))
1604                 goto regen;
1605         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1606                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1607                         goto regen;
1608                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1609                         goto regen;
1610         }
1611
1612         return 0;
1613 }
1614
1615 static void ipv6_regen_rndid(unsigned long data)
1616 {
1617         struct inet6_dev *idev = (struct inet6_dev *) data;
1618         unsigned long expires;
1619
1620         rcu_read_lock_bh();
1621         write_lock_bh(&idev->lock);
1622
1623         if (idev->dead)
1624                 goto out;
1625
1626         if (__ipv6_regen_rndid(idev) < 0)
1627                 goto out;
1628
1629         expires = jiffies +
1630                 idev->cnf.temp_prefered_lft * HZ -
1631                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1632         if (time_before(expires, jiffies)) {
1633                 printk(KERN_WARNING
1634                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1635                         idev->dev->name);
1636                 goto out;
1637         }
1638
1639         if (!mod_timer(&idev->regen_timer, expires))
1640                 in6_dev_hold(idev);
1641
1642 out:
1643         write_unlock_bh(&idev->lock);
1644         rcu_read_unlock_bh();
1645         in6_dev_put(idev);
1646 }
1647
1648 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1649         int ret = 0;
1650
1651         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1652                 ret = __ipv6_regen_rndid(idev);
1653         return ret;
1654 }
1655 #endif
1656
1657 /*
1658  *      Add prefix route.
1659  */
1660
1661 static void
1662 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1663                       unsigned long expires, u32 flags)
1664 {
1665         struct fib6_config cfg = {
1666                 .fc_table = RT6_TABLE_PREFIX,
1667                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1668                 .fc_ifindex = dev->ifindex,
1669                 .fc_expires = expires,
1670                 .fc_dst_len = plen,
1671                 .fc_flags = RTF_UP | flags,
1672                 .fc_nlinfo.nl_net = dev_net(dev),
1673                 .fc_protocol = RTPROT_KERNEL,
1674         };
1675
1676         ipv6_addr_copy(&cfg.fc_dst, pfx);
1677
1678         /* Prevent useless cloning on PtP SIT.
1679            This thing is done here expecting that the whole
1680            class of non-broadcast devices need not cloning.
1681          */
1682 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1683         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1684                 cfg.fc_flags |= RTF_NONEXTHOP;
1685 #endif
1686
1687         ip6_route_add(&cfg);
1688 }
1689
1690 /* Create "default" multicast route to the interface */
1691
1692 static void addrconf_add_mroute(struct net_device *dev)
1693 {
1694         struct fib6_config cfg = {
1695                 .fc_table = RT6_TABLE_LOCAL,
1696                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1697                 .fc_ifindex = dev->ifindex,
1698                 .fc_dst_len = 8,
1699                 .fc_flags = RTF_UP,
1700                 .fc_nlinfo.nl_net = dev_net(dev),
1701         };
1702
1703         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1704
1705         ip6_route_add(&cfg);
1706 }
1707
1708 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1709 static void sit_route_add(struct net_device *dev)
1710 {
1711         struct fib6_config cfg = {
1712                 .fc_table = RT6_TABLE_MAIN,
1713                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1714                 .fc_ifindex = dev->ifindex,
1715                 .fc_dst_len = 96,
1716                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1717                 .fc_nlinfo.nl_net = dev_net(dev),
1718         };
1719
1720         /* prefix length - 96 bits "::d.d.d.d" */
1721         ip6_route_add(&cfg);
1722 }
1723 #endif
1724
1725 static void addrconf_add_lroute(struct net_device *dev)
1726 {
1727         struct in6_addr addr;
1728
1729         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1730         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1731 }
1732
1733 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1734 {
1735         struct inet6_dev *idev;
1736
1737         ASSERT_RTNL();
1738
1739         if ((idev = ipv6_find_idev(dev)) == NULL)
1740                 return NULL;
1741
1742         /* Add default multicast route */
1743         addrconf_add_mroute(dev);
1744
1745         /* Add link local route */
1746         addrconf_add_lroute(dev);
1747         return idev;
1748 }
1749
1750 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1751 {
1752         struct prefix_info *pinfo;
1753         __u32 valid_lft;
1754         __u32 prefered_lft;
1755         int addr_type;
1756         struct inet6_dev *in6_dev;
1757         struct net *net = dev_net(dev);
1758
1759         pinfo = (struct prefix_info *) opt;
1760
1761         if (len < sizeof(struct prefix_info)) {
1762                 ADBG(("addrconf: prefix option too short\n"));
1763                 return;
1764         }
1765
1766         /*
1767          *      Validation checks ([ADDRCONF], page 19)
1768          */
1769
1770         addr_type = ipv6_addr_type(&pinfo->prefix);
1771
1772         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1773                 return;
1774
1775         valid_lft = ntohl(pinfo->valid);
1776         prefered_lft = ntohl(pinfo->prefered);
1777
1778         if (prefered_lft > valid_lft) {
1779                 if (net_ratelimit())
1780                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1781                 return;
1782         }
1783
1784         in6_dev = in6_dev_get(dev);
1785
1786         if (in6_dev == NULL) {
1787                 if (net_ratelimit())
1788                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1789                 return;
1790         }
1791
1792         /*
1793          *      Two things going on here:
1794          *      1) Add routes for on-link prefixes
1795          *      2) Configure prefixes with the auto flag set
1796          */
1797
1798         if (pinfo->onlink) {
1799                 struct rt6_info *rt;
1800                 unsigned long rt_expires;
1801
1802                 /* Avoid arithmetic overflow. Really, we could
1803                  * save rt_expires in seconds, likely valid_lft,
1804                  * but it would require division in fib gc, that it
1805                  * not good.
1806                  */
1807                 if (HZ > USER_HZ)
1808                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1809                 else
1810                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1811
1812                 if (addrconf_finite_timeout(rt_expires))
1813                         rt_expires *= HZ;
1814
1815                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1816                                 dev->ifindex, 1);
1817
1818                 if (rt && addrconf_is_prefix_route(rt)) {
1819                         /* Autoconf prefix route */
1820                         if (valid_lft == 0) {
1821                                 ip6_del_rt(rt);
1822                                 rt = NULL;
1823                         } else if (addrconf_finite_timeout(rt_expires)) {
1824                                 /* not infinity */
1825                                 rt->rt6i_expires = jiffies + rt_expires;
1826                                 rt->rt6i_flags |= RTF_EXPIRES;
1827                         } else {
1828                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1829                                 rt->rt6i_expires = 0;
1830                         }
1831                 } else if (valid_lft) {
1832                         clock_t expires = 0;
1833                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1834                         if (addrconf_finite_timeout(rt_expires)) {
1835                                 /* not infinity */
1836                                 flags |= RTF_EXPIRES;
1837                                 expires = jiffies_to_clock_t(rt_expires);
1838                         }
1839                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1840                                               dev, expires, flags);
1841                 }
1842                 if (rt)
1843                         dst_release(&rt->u.dst);
1844         }
1845
1846         /* Try to figure out our local address for this prefix */
1847
1848         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1849                 struct inet6_ifaddr * ifp;
1850                 struct in6_addr addr;
1851                 int create = 0, update_lft = 0;
1852
1853                 if (pinfo->prefix_len == 64) {
1854                         memcpy(&addr, &pinfo->prefix, 8);
1855                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1856                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1857                                 in6_dev_put(in6_dev);
1858                                 return;
1859                         }
1860                         goto ok;
1861                 }
1862                 if (net_ratelimit())
1863                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1864                                pinfo->prefix_len);
1865                 in6_dev_put(in6_dev);
1866                 return;
1867
1868 ok:
1869
1870                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1871
1872                 if (ifp == NULL && valid_lft) {
1873                         int max_addresses = in6_dev->cnf.max_addresses;
1874                         u32 addr_flags = 0;
1875
1876 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1877                         if (in6_dev->cnf.optimistic_dad &&
1878                             !net->ipv6.devconf_all->forwarding)
1879                                 addr_flags = IFA_F_OPTIMISTIC;
1880 #endif
1881
1882                         /* Do not allow to create too much of autoconfigured
1883                          * addresses; this would be too easy way to crash kernel.
1884                          */
1885                         if (!max_addresses ||
1886                             ipv6_count_addresses(in6_dev) < max_addresses)
1887                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1888                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1889                                                     addr_flags);
1890
1891                         if (!ifp || IS_ERR(ifp)) {
1892                                 in6_dev_put(in6_dev);
1893                                 return;
1894                         }
1895
1896                         update_lft = create = 1;
1897                         ifp->cstamp = jiffies;
1898                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1899                 }
1900
1901                 if (ifp) {
1902                         int flags;
1903                         unsigned long now;
1904 #ifdef CONFIG_IPV6_PRIVACY
1905                         struct inet6_ifaddr *ift;
1906 #endif
1907                         u32 stored_lft;
1908
1909                         /* update lifetime (RFC2462 5.5.3 e) */
1910                         spin_lock(&ifp->lock);
1911                         now = jiffies;
1912                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1913                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1914                         else
1915                                 stored_lft = 0;
1916                         if (!update_lft && stored_lft) {
1917                                 if (valid_lft > MIN_VALID_LIFETIME ||
1918                                     valid_lft > stored_lft)
1919                                         update_lft = 1;
1920                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1921                                         /* valid_lft <= stored_lft is always true */
1922                                         /*
1923                                          * RFC 4862 Section 5.5.3e:
1924                                          * "Note that the preferred lifetime of
1925                                          *  the corresponding address is always
1926                                          *  reset to the Preferred Lifetime in
1927                                          *  the received Prefix Information
1928                                          *  option, regardless of whether the
1929                                          *  valid lifetime is also reset or
1930                                          *  ignored."
1931                                          *
1932                                          *  So if the preferred lifetime in
1933                                          *  this advertisement is different
1934                                          *  than what we have stored, but the
1935                                          *  valid lifetime is invalid, just
1936                                          *  reset prefered_lft.
1937                                          *
1938                                          *  We must set the valid lifetime
1939                                          *  to the stored lifetime since we'll
1940                                          *  be updating the timestamp below,
1941                                          *  else we'll set it back to the
1942                                          *  minumum.
1943                                          */
1944                                         if (prefered_lft != ifp->prefered_lft) {
1945                                                 valid_lft = stored_lft;
1946                                                 update_lft = 1;
1947                                         }
1948                                 } else {
1949                                         valid_lft = MIN_VALID_LIFETIME;
1950                                         if (valid_lft < prefered_lft)
1951                                                 prefered_lft = valid_lft;
1952                                         update_lft = 1;
1953                                 }
1954                         }
1955
1956                         if (update_lft) {
1957                                 ifp->valid_lft = valid_lft;
1958                                 ifp->prefered_lft = prefered_lft;
1959                                 ifp->tstamp = now;
1960                                 flags = ifp->flags;
1961                                 ifp->flags &= ~IFA_F_DEPRECATED;
1962                                 spin_unlock(&ifp->lock);
1963
1964                                 if (!(flags&IFA_F_TENTATIVE))
1965                                         ipv6_ifa_notify(0, ifp);
1966                         } else
1967                                 spin_unlock(&ifp->lock);
1968
1969 #ifdef CONFIG_IPV6_PRIVACY
1970                         read_lock_bh(&in6_dev->lock);
1971                         /* update all temporary addresses in the list */
1972                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1973                                 /*
1974                                  * When adjusting the lifetimes of an existing
1975                                  * temporary address, only lower the lifetimes.
1976                                  * Implementations must not increase the
1977                                  * lifetimes of an existing temporary address
1978                                  * when processing a Prefix Information Option.
1979                                  */
1980                                 if (ifp != ift->ifpub)
1981                                         continue;
1982
1983                                 spin_lock(&ift->lock);
1984                                 flags = ift->flags;
1985                                 if (ift->valid_lft > valid_lft &&
1986                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1987                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1988                                 if (ift->prefered_lft > prefered_lft &&
1989                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1990                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1991                                 spin_unlock(&ift->lock);
1992                                 if (!(flags&IFA_F_TENTATIVE))
1993                                         ipv6_ifa_notify(0, ift);
1994                         }
1995
1996                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1997                                 /*
1998                                  * When a new public address is created as described in [ADDRCONF],
1999                                  * also create a new temporary address.
2000                                  */
2001                                 read_unlock_bh(&in6_dev->lock);
2002                                 ipv6_create_tempaddr(ifp, NULL);
2003                         } else {
2004                                 read_unlock_bh(&in6_dev->lock);
2005                         }
2006 #endif
2007                         in6_ifa_put(ifp);
2008                         addrconf_verify(0);
2009                 }
2010         }
2011         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2012         in6_dev_put(in6_dev);
2013 }
2014
2015 /*
2016  *      Set destination address.
2017  *      Special case for SIT interfaces where we create a new "virtual"
2018  *      device.
2019  */
2020 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2021 {
2022         struct in6_ifreq ireq;
2023         struct net_device *dev;
2024         int err = -EINVAL;
2025
2026         rtnl_lock();
2027
2028         err = -EFAULT;
2029         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2030                 goto err_exit;
2031
2032         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2033
2034         err = -ENODEV;
2035         if (dev == NULL)
2036                 goto err_exit;
2037
2038 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2039         if (dev->type == ARPHRD_SIT) {
2040                 const struct net_device_ops *ops = dev->netdev_ops;
2041                 struct ifreq ifr;
2042                 struct ip_tunnel_parm p;
2043
2044                 err = -EADDRNOTAVAIL;
2045                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2046                         goto err_exit;
2047
2048                 memset(&p, 0, sizeof(p));
2049                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2050                 p.iph.saddr = 0;
2051                 p.iph.version = 4;
2052                 p.iph.ihl = 5;
2053                 p.iph.protocol = IPPROTO_IPV6;
2054                 p.iph.ttl = 64;
2055                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2056
2057                 if (ops->ndo_do_ioctl) {
2058                         mm_segment_t oldfs = get_fs();
2059
2060                         set_fs(KERNEL_DS);
2061                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2062                         set_fs(oldfs);
2063                 } else
2064                         err = -EOPNOTSUPP;
2065
2066                 if (err == 0) {
2067                         err = -ENOBUFS;
2068                         dev = __dev_get_by_name(net, p.name);
2069                         if (!dev)
2070                                 goto err_exit;
2071                         err = dev_open(dev);
2072                 }
2073         }
2074 #endif
2075
2076 err_exit:
2077         rtnl_unlock();
2078         return err;
2079 }
2080
2081 /*
2082  *      Manual configuration of address on an interface
2083  */
2084 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2085                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2086                           __u32 valid_lft)
2087 {
2088         struct inet6_ifaddr *ifp;
2089         struct inet6_dev *idev;
2090         struct net_device *dev;
2091         int scope;
2092         u32 flags;
2093         clock_t expires;
2094         unsigned long timeout;
2095
2096         ASSERT_RTNL();
2097
2098         if (plen > 128)
2099                 return -EINVAL;
2100
2101         /* check the lifetime */
2102         if (!valid_lft || prefered_lft > valid_lft)
2103                 return -EINVAL;
2104
2105         dev = __dev_get_by_index(net, ifindex);
2106         if (!dev)
2107                 return -ENODEV;
2108
2109         if ((idev = addrconf_add_dev(dev)) == NULL)
2110                 return -ENOBUFS;
2111
2112         scope = ipv6_addr_scope(pfx);
2113
2114         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2115         if (addrconf_finite_timeout(timeout)) {
2116                 expires = jiffies_to_clock_t(timeout * HZ);
2117                 valid_lft = timeout;
2118                 flags = RTF_EXPIRES;
2119         } else {
2120                 expires = 0;
2121                 flags = 0;
2122                 ifa_flags |= IFA_F_PERMANENT;
2123         }
2124
2125         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2126         if (addrconf_finite_timeout(timeout)) {
2127                 if (timeout == 0)
2128                         ifa_flags |= IFA_F_DEPRECATED;
2129                 prefered_lft = timeout;
2130         }
2131
2132         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2133
2134         if (!IS_ERR(ifp)) {
2135                 spin_lock_bh(&ifp->lock);
2136                 ifp->valid_lft = valid_lft;
2137                 ifp->prefered_lft = prefered_lft;
2138                 ifp->tstamp = jiffies;
2139                 spin_unlock_bh(&ifp->lock);
2140
2141                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2142                                       expires, flags);
2143                 /*
2144                  * Note that section 3.1 of RFC 4429 indicates
2145                  * that the Optimistic flag should not be set for
2146                  * manually configured addresses
2147                  */
2148                 addrconf_dad_start(ifp, 0);
2149                 in6_ifa_put(ifp);
2150                 addrconf_verify(0);
2151                 return 0;
2152         }
2153
2154         return PTR_ERR(ifp);
2155 }
2156
2157 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2158                           unsigned int plen)
2159 {
2160         struct inet6_ifaddr *ifp;
2161         struct inet6_dev *idev;
2162         struct net_device *dev;
2163
2164         if (plen > 128)
2165                 return -EINVAL;
2166
2167         dev = __dev_get_by_index(net, ifindex);
2168         if (!dev)
2169                 return -ENODEV;
2170
2171         if ((idev = __in6_dev_get(dev)) == NULL)
2172                 return -ENXIO;
2173
2174         read_lock_bh(&idev->lock);
2175         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2176                 if (ifp->prefix_len == plen &&
2177                     ipv6_addr_equal(pfx, &ifp->addr)) {
2178                         in6_ifa_hold(ifp);
2179                         read_unlock_bh(&idev->lock);
2180
2181                         ipv6_del_addr(ifp);
2182
2183                         /* If the last address is deleted administratively,
2184                            disable IPv6 on this interface.
2185                          */
2186                         if (idev->addr_list == NULL)
2187                                 addrconf_ifdown(idev->dev, 1);
2188                         return 0;
2189                 }
2190         }
2191         read_unlock_bh(&idev->lock);
2192         return -EADDRNOTAVAIL;
2193 }
2194
2195
2196 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2197 {
2198         struct in6_ifreq ireq;
2199         int err;
2200
2201         if (!capable(CAP_NET_ADMIN))
2202                 return -EPERM;
2203
2204         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2205                 return -EFAULT;
2206
2207         rtnl_lock();
2208         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2209                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2210                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2211         rtnl_unlock();
2212         return err;
2213 }
2214
2215 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2216 {
2217         struct in6_ifreq ireq;
2218         int err;
2219
2220         if (!capable(CAP_NET_ADMIN))
2221                 return -EPERM;
2222
2223         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2224                 return -EFAULT;
2225
2226         rtnl_lock();
2227         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2228                              ireq.ifr6_prefixlen);
2229         rtnl_unlock();
2230         return err;
2231 }
2232
2233 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2234                      int plen, int scope)
2235 {
2236         struct inet6_ifaddr *ifp;
2237
2238         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2239         if (!IS_ERR(ifp)) {
2240                 spin_lock_bh(&ifp->lock);
2241                 ifp->flags &= ~IFA_F_TENTATIVE;
2242                 spin_unlock_bh(&ifp->lock);
2243                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2244                 in6_ifa_put(ifp);
2245         }
2246 }
2247
2248 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2249 static void sit_add_v4_addrs(struct inet6_dev *idev)
2250 {
2251         struct in6_addr addr;
2252         struct net_device *dev;
2253         struct net *net = dev_net(idev->dev);
2254         int scope;
2255
2256         ASSERT_RTNL();
2257
2258         memset(&addr, 0, sizeof(struct in6_addr));
2259         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2260
2261         if (idev->dev->flags&IFF_POINTOPOINT) {
2262                 addr.s6_addr32[0] = htonl(0xfe800000);
2263                 scope = IFA_LINK;
2264         } else {
2265                 scope = IPV6_ADDR_COMPATv4;
2266         }
2267
2268         if (addr.s6_addr32[3]) {
2269                 add_addr(idev, &addr, 128, scope);
2270                 return;
2271         }
2272
2273         for_each_netdev(net, dev) {
2274                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2275                 if (in_dev && (dev->flags & IFF_UP)) {
2276                         struct in_ifaddr * ifa;
2277
2278                         int flag = scope;
2279
2280                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2281                                 int plen;
2282
2283                                 addr.s6_addr32[3] = ifa->ifa_local;
2284
2285                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2286                                         continue;
2287                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2288                                         if (idev->dev->flags&IFF_POINTOPOINT)
2289                                                 continue;
2290                                         flag |= IFA_HOST;
2291                                 }
2292                                 if (idev->dev->flags&IFF_POINTOPOINT)
2293                                         plen = 64;
2294                                 else
2295                                         plen = 96;
2296
2297                                 add_addr(idev, &addr, plen, flag);
2298                         }
2299                 }
2300         }
2301 }
2302 #endif
2303
2304 static void init_loopback(struct net_device *dev)
2305 {
2306         struct inet6_dev  *idev;
2307
2308         /* ::1 */
2309
2310         ASSERT_RTNL();
2311
2312         if ((idev = ipv6_find_idev(dev)) == NULL) {
2313                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2314                 return;
2315         }
2316
2317         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2318 }
2319
2320 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2321 {
2322         struct inet6_ifaddr * ifp;
2323         u32 addr_flags = IFA_F_PERMANENT;
2324
2325 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2326         if (idev->cnf.optimistic_dad &&
2327             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2328                 addr_flags |= IFA_F_OPTIMISTIC;
2329 #endif
2330
2331
2332         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2333         if (!IS_ERR(ifp)) {
2334                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2335                 addrconf_dad_start(ifp, 0);
2336                 in6_ifa_put(ifp);
2337         }
2338 }
2339
2340 static void addrconf_dev_config(struct net_device *dev)
2341 {
2342         struct in6_addr addr;
2343         struct inet6_dev    * idev;
2344
2345         ASSERT_RTNL();
2346
2347         if ((dev->type != ARPHRD_ETHER) &&
2348             (dev->type != ARPHRD_FDDI) &&
2349             (dev->type != ARPHRD_IEEE802_TR) &&
2350             (dev->type != ARPHRD_ARCNET) &&
2351             (dev->type != ARPHRD_INFINIBAND)) {
2352                 /* Alas, we support only Ethernet autoconfiguration. */
2353                 return;
2354         }
2355
2356         idev = addrconf_add_dev(dev);
2357         if (idev == NULL)
2358                 return;
2359
2360         memset(&addr, 0, sizeof(struct in6_addr));
2361         addr.s6_addr32[0] = htonl(0xFE800000);
2362
2363         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2364                 addrconf_add_linklocal(idev, &addr);
2365 }
2366
2367 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2368 static void addrconf_sit_config(struct net_device *dev)
2369 {
2370         struct inet6_dev *idev;
2371
2372         ASSERT_RTNL();
2373
2374         /*
2375          * Configure the tunnel with one of our IPv4
2376          * addresses... we should configure all of
2377          * our v4 addrs in the tunnel
2378          */
2379
2380         if ((idev = ipv6_find_idev(dev)) == NULL) {
2381                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2382                 return;
2383         }
2384
2385         if (dev->priv_flags & IFF_ISATAP) {
2386                 struct in6_addr addr;
2387
2388                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2389                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2390                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2391                         addrconf_add_linklocal(idev, &addr);
2392                 return;
2393         }
2394
2395         sit_add_v4_addrs(idev);
2396
2397         if (dev->flags&IFF_POINTOPOINT) {
2398                 addrconf_add_mroute(dev);
2399                 addrconf_add_lroute(dev);
2400         } else
2401                 sit_route_add(dev);
2402 }
2403 #endif
2404
2405 static inline int
2406 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2407 {
2408         struct in6_addr lladdr;
2409
2410         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2411                 addrconf_add_linklocal(idev, &lladdr);
2412                 return 0;
2413         }
2414         return -1;
2415 }
2416
2417 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2418 {
2419         struct net_device *link_dev;
2420         struct net *net = dev_net(idev->dev);
2421
2422         /* first try to inherit the link-local address from the link device */
2423         if (idev->dev->iflink &&
2424             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2425                 if (!ipv6_inherit_linklocal(idev, link_dev))
2426                         return;
2427         }
2428         /* then try to inherit it from any device */
2429         for_each_netdev(net, link_dev) {
2430                 if (!ipv6_inherit_linklocal(idev, link_dev))
2431                         return;
2432         }
2433         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2434 }
2435
2436 /*
2437  * Autoconfigure tunnel with a link-local address so routing protocols,
2438  * DHCPv6, MLD etc. can be run over the virtual link
2439  */
2440
2441 static void addrconf_ip6_tnl_config(struct net_device *dev)
2442 {
2443         struct inet6_dev *idev;
2444
2445         ASSERT_RTNL();
2446
2447         if ((idev = addrconf_add_dev(dev)) == NULL) {
2448                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2449                 return;
2450         }
2451         ip6_tnl_add_linklocal(idev);
2452 }
2453
2454 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2455                            void * data)
2456 {
2457         struct net_device *dev = (struct net_device *) data;
2458         struct inet6_dev *idev = __in6_dev_get(dev);
2459         int run_pending = 0;
2460         int err;
2461
2462         switch(event) {
2463         case NETDEV_REGISTER:
2464                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2465                         idev = ipv6_add_dev(dev);
2466                         if (!idev)
2467                                 return notifier_from_errno(-ENOMEM);
2468                 }
2469                 break;
2470         case NETDEV_UP:
2471         case NETDEV_CHANGE:
2472                 if (dev->flags & IFF_SLAVE)
2473                         break;
2474
2475                 if (event == NETDEV_UP) {
2476                         if (!addrconf_qdisc_ok(dev)) {
2477                                 /* device is not ready yet. */
2478                                 printk(KERN_INFO
2479                                         "ADDRCONF(NETDEV_UP): %s: "
2480                                         "link is not ready\n",
2481                                         dev->name);
2482                                 break;
2483                         }
2484
2485                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2486                                 idev = ipv6_add_dev(dev);
2487
2488                         if (idev) {
2489                                 idev->if_flags |= IF_READY;
2490                                 run_pending = 1;
2491                         }
2492                 } else {
2493                         if (!addrconf_qdisc_ok(dev)) {
2494                                 /* device is still not ready. */
2495                                 break;
2496                         }
2497
2498                         if (idev) {
2499                                 if (idev->if_flags & IF_READY) {
2500                                         /* device is already configured. */
2501                                         break;
2502                                 }
2503                                 idev->if_flags |= IF_READY;
2504                         }
2505
2506                         printk(KERN_INFO
2507                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2508                                         "link becomes ready\n",
2509                                         dev->name);
2510
2511                         run_pending = 1;
2512                 }
2513
2514                 switch(dev->type) {
2515 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2516                 case ARPHRD_SIT:
2517                         addrconf_sit_config(dev);
2518                         break;
2519 #endif
2520                 case ARPHRD_TUNNEL6:
2521                         addrconf_ip6_tnl_config(dev);
2522                         break;
2523                 case ARPHRD_LOOPBACK:
2524                         init_loopback(dev);
2525                         break;
2526
2527                 default:
2528                         addrconf_dev_config(dev);
2529                         break;
2530                 }
2531                 if (idev) {
2532                         if (run_pending)
2533                                 addrconf_dad_run(idev);
2534
2535                         /* If the MTU changed during the interface down, when the
2536                            interface up, the changed MTU must be reflected in the
2537                            idev as well as routers.
2538                          */
2539                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2540                                 rt6_mtu_change(dev, dev->mtu);
2541                                 idev->cnf.mtu6 = dev->mtu;
2542                         }
2543                         idev->tstamp = jiffies;
2544                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2545                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2546                            stop IPv6 on this interface.
2547                          */
2548                         if (dev->mtu < IPV6_MIN_MTU)
2549                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2550                 }
2551                 break;
2552
2553         case NETDEV_CHANGEMTU:
2554                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2555                         rt6_mtu_change(dev, dev->mtu);
2556                         idev->cnf.mtu6 = dev->mtu;
2557                         break;
2558                 }
2559
2560                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2561                         idev = ipv6_add_dev(dev);
2562                         if (idev)
2563                                 break;
2564                 }
2565
2566                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2567
2568         case NETDEV_DOWN:
2569         case NETDEV_UNREGISTER:
2570                 /*
2571                  *      Remove all addresses from this interface.
2572                  */
2573                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2574                 break;
2575
2576         case NETDEV_CHANGENAME:
2577                 if (idev) {
2578                         snmp6_unregister_dev(idev);
2579                         addrconf_sysctl_unregister(idev);
2580                         addrconf_sysctl_register(idev);
2581                         err = snmp6_register_dev(idev);
2582                         if (err)
2583                                 return notifier_from_errno(err);
2584                 }
2585                 break;
2586         case NETDEV_BONDING_OLDTYPE:
2587         case NETDEV_BONDING_NEWTYPE:
2588                 addrconf_bonding_change(dev, event);
2589                 break;
2590         }
2591
2592         return NOTIFY_OK;
2593 }
2594
2595 /*
2596  *      addrconf module should be notified of a device going up
2597  */
2598 static struct notifier_block ipv6_dev_notf = {
2599         .notifier_call = addrconf_notify,
2600         .priority = 0
2601 };
2602
2603 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2604 {
2605         struct inet6_dev *idev;
2606         ASSERT_RTNL();
2607
2608         idev = __in6_dev_get(dev);
2609
2610         if (event == NETDEV_BONDING_NEWTYPE)
2611                 ipv6_mc_remap(idev);
2612         else if (event == NETDEV_BONDING_OLDTYPE)
2613                 ipv6_mc_unmap(idev);
2614 }
2615
2616 static int addrconf_ifdown(struct net_device *dev, int how)
2617 {
2618         struct inet6_dev *idev;
2619         struct inet6_ifaddr *ifa, **bifa;
2620         struct net *net = dev_net(dev);
2621         int i;
2622
2623         ASSERT_RTNL();
2624
2625         rt6_ifdown(net, dev);
2626         neigh_ifdown(&nd_tbl, dev);
2627
2628         idev = __in6_dev_get(dev);
2629         if (idev == NULL)
2630                 return -ENODEV;
2631
2632         /* Step 1: remove reference to ipv6 device from parent device.
2633                    Do not dev_put!
2634          */
2635         if (how) {
2636                 idev->dead = 1;
2637
2638                 /* protected by rtnl_lock */
2639                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2640
2641                 /* Step 1.5: remove snmp6 entry */
2642                 snmp6_unregister_dev(idev);
2643
2644         }
2645
2646         /* Step 2: clear hash table */
2647         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2648                 bifa = &inet6_addr_lst[i];
2649
2650                 write_lock_bh(&addrconf_hash_lock);
2651                 while ((ifa = *bifa) != NULL) {
2652                         if (ifa->idev == idev &&
2653                             (how || !(ifa->flags&IFA_F_PERMANENT))) {
2654                                 *bifa = ifa->lst_next;
2655                                 ifa->lst_next = NULL;
2656                                 addrconf_del_timer(ifa);
2657                                 in6_ifa_put(ifa);
2658                                 continue;
2659                         }
2660                         bifa = &ifa->lst_next;
2661                 }
2662                 write_unlock_bh(&addrconf_hash_lock);
2663         }
2664
2665         write_lock_bh(&idev->lock);
2666
2667         /* Step 3: clear flags for stateless addrconf */
2668         if (!how)
2669                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2670
2671         /* Step 4: clear address list */
2672 #ifdef CONFIG_IPV6_PRIVACY
2673         if (how && del_timer(&idev->regen_timer))
2674                 in6_dev_put(idev);
2675
2676         /* clear tempaddr list */
2677         while ((ifa = idev->tempaddr_list) != NULL) {
2678                 idev->tempaddr_list = ifa->tmp_next;
2679                 ifa->tmp_next = NULL;
2680                 ifa->dead = 1;
2681                 write_unlock_bh(&idev->lock);
2682                 spin_lock_bh(&ifa->lock);
2683
2684                 if (ifa->ifpub) {
2685                         in6_ifa_put(ifa->ifpub);
2686                         ifa->ifpub = NULL;
2687                 }
2688                 spin_unlock_bh(&ifa->lock);
2689                 in6_ifa_put(ifa);
2690                 write_lock_bh(&idev->lock);
2691         }
2692 #endif
2693         bifa = &idev->addr_list;
2694         while ((ifa = *bifa) != NULL) {
2695                 if (how == 0 && (ifa->flags&IFA_F_PERMANENT)) {
2696                         /* Retain permanent address on admin down */
2697                         bifa = &ifa->if_next;
2698
2699                         /* Restart DAD if needed when link comes back up */
2700                         if ( !((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2701                                idev->cnf.accept_dad <= 0 ||
2702                                (ifa->flags & IFA_F_NODAD)))
2703                                 ifa->flags |= IFA_F_TENTATIVE;
2704                 } else {
2705                         *bifa = ifa->if_next;
2706                         ifa->if_next = NULL;
2707
2708                         ifa->dead = 1;
2709                         write_unlock_bh(&idev->lock);
2710
2711                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2712                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2713                         in6_ifa_put(ifa);
2714
2715                         write_lock_bh(&idev->lock);
2716                 }
2717         }
2718         write_unlock_bh(&idev->lock);
2719
2720         /* Step 5: Discard multicast list */
2721
2722         if (how)
2723                 ipv6_mc_destroy_dev(idev);
2724         else
2725                 ipv6_mc_down(idev);
2726
2727         idev->tstamp = jiffies;
2728
2729         /* Shot the device (if unregistered) */
2730
2731         if (how) {
2732                 addrconf_sysctl_unregister(idev);
2733                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2734                 neigh_ifdown(&nd_tbl, dev);
2735                 in6_dev_put(idev);
2736         }
2737         return 0;
2738 }
2739
2740 static void addrconf_rs_timer(unsigned long data)
2741 {
2742         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2743
2744         if (ifp->idev->cnf.forwarding)
2745                 goto out;
2746
2747         if (ifp->idev->if_flags & IF_RA_RCVD) {
2748                 /*
2749                  *      Announcement received after solicitation
2750                  *      was sent
2751                  */
2752                 goto out;
2753         }
2754
2755         spin_lock(&ifp->lock);
2756         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2757                 /* The wait after the last probe can be shorter */
2758                 addrconf_mod_timer(ifp, AC_RS,
2759                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2760                                    ifp->idev->cnf.rtr_solicit_delay :
2761                                    ifp->idev->cnf.rtr_solicit_interval);
2762                 spin_unlock(&ifp->lock);
2763
2764                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2765         } else {
2766                 spin_unlock(&ifp->lock);
2767                 /*
2768                  * Note: we do not support deprecated "all on-link"
2769                  * assumption any longer.
2770                  */
2771                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2772                        ifp->idev->dev->name);
2773         }
2774
2775 out:
2776         in6_ifa_put(ifp);
2777 }
2778
2779 /*
2780  *      Duplicate Address Detection
2781  */
2782 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2783 {
2784         unsigned long rand_num;
2785         struct inet6_dev *idev = ifp->idev;
2786
2787         if (ifp->flags & IFA_F_OPTIMISTIC)
2788                 rand_num = 0;
2789         else
2790                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2791
2792         ifp->probes = idev->cnf.dad_transmits;
2793         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2794 }
2795
2796 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2797 {
2798         struct inet6_dev *idev = ifp->idev;
2799         struct net_device *dev = idev->dev;
2800
2801         addrconf_join_solict(dev, &ifp->addr);
2802
2803         net_srandom(ifp->addr.s6_addr32[3]);
2804
2805         read_lock_bh(&idev->lock);
2806         if (ifp->dead)
2807                 goto out;
2808
2809         spin_lock(&ifp->lock);
2810         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2811             idev->cnf.accept_dad < 1 ||
2812             !(ifp->flags&IFA_F_TENTATIVE) ||
2813             ifp->flags & IFA_F_NODAD) {
2814                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2815                 spin_unlock(&ifp->lock);
2816                 read_unlock_bh(&idev->lock);
2817
2818                 addrconf_dad_completed(ifp);
2819                 return;
2820         }
2821
2822         if (!(idev->if_flags & IF_READY)) {
2823                 spin_unlock(&ifp->lock);
2824                 read_unlock_bh(&idev->lock);
2825                 /*
2826                  * If the device is not ready:
2827                  * - keep it tentative if it is a permanent address.
2828                  * - otherwise, kill it.
2829                  */
2830                 in6_ifa_hold(ifp);
2831                 addrconf_dad_stop(ifp, 0);
2832                 return;
2833         }
2834
2835         /*
2836          * Optimistic nodes can start receiving
2837          * Frames right away
2838          */
2839         if(ifp->flags & IFA_F_OPTIMISTIC)
2840                 ip6_ins_rt(ifp->rt);
2841
2842         addrconf_dad_kick(ifp);
2843         spin_unlock(&ifp->lock);
2844 out:
2845         read_unlock_bh(&idev->lock);
2846 }
2847
2848 static void addrconf_dad_timer(unsigned long data)
2849 {
2850         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2851         struct inet6_dev *idev = ifp->idev;
2852         struct in6_addr mcaddr;
2853
2854         read_lock_bh(&idev->lock);
2855         if (idev->dead) {
2856                 read_unlock_bh(&idev->lock);
2857                 goto out;
2858         }
2859
2860         spin_lock(&ifp->lock);
2861         if (ifp->probes == 0) {
2862                 /*
2863                  * DAD was successful
2864                  */
2865
2866                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2867                 spin_unlock(&ifp->lock);
2868                 read_unlock_bh(&idev->lock);
2869
2870                 addrconf_dad_completed(ifp);
2871
2872                 goto out;
2873         }
2874
2875         ifp->probes--;
2876         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2877         spin_unlock(&ifp->lock);
2878         read_unlock_bh(&idev->lock);
2879
2880         /* send a neighbour solicitation for our addr */
2881         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2882         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2883 out:
2884         in6_ifa_put(ifp);
2885 }
2886
2887 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2888 {
2889         struct net_device *     dev = ifp->idev->dev;
2890
2891         /*
2892          *      Configure the address for reception. Now it is valid.
2893          */
2894
2895         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2896
2897         /* If added prefix is link local and forwarding is off,
2898            start sending router solicitations.
2899          */
2900
2901         if (ifp->idev->cnf.forwarding == 0 &&
2902             ifp->idev->cnf.rtr_solicits > 0 &&
2903             (dev->flags&IFF_LOOPBACK) == 0 &&
2904             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2905                 /*
2906                  *      If a host as already performed a random delay
2907                  *      [...] as part of DAD [...] there is no need
2908                  *      to delay again before sending the first RS
2909                  */
2910                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2911
2912                 spin_lock_bh(&ifp->lock);
2913                 ifp->probes = 1;
2914                 ifp->idev->if_flags |= IF_RS_SENT;
2915                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2916                 spin_unlock_bh(&ifp->lock);
2917         }
2918 }
2919
2920 static void addrconf_dad_run(struct inet6_dev *idev) {
2921         struct inet6_ifaddr *ifp;
2922
2923         read_lock_bh(&idev->lock);
2924         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2925                 spin_lock(&ifp->lock);
2926                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2927                         spin_unlock(&ifp->lock);
2928                         continue;
2929                 }
2930                 spin_unlock(&ifp->lock);
2931                 addrconf_dad_kick(ifp);
2932         }
2933         read_unlock_bh(&idev->lock);
2934 }
2935
2936 #ifdef CONFIG_PROC_FS
2937 struct if6_iter_state {
2938         struct seq_net_private p;
2939         int bucket;
2940 };
2941
2942 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2943 {
2944         struct inet6_ifaddr *ifa = NULL;
2945         struct if6_iter_state *state = seq->private;
2946         struct net *net = seq_file_net(seq);
2947
2948         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2949                 ifa = inet6_addr_lst[state->bucket];
2950
2951                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2952                         ifa = ifa->lst_next;
2953                 if (ifa)
2954                         break;
2955         }
2956         return ifa;
2957 }
2958
2959 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2960 {
2961         struct if6_iter_state *state = seq->private;
2962         struct net *net = seq_file_net(seq);
2963
2964         ifa = ifa->lst_next;
2965 try_again:
2966         if (ifa) {
2967                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2968                         ifa = ifa->lst_next;
2969                         goto try_again;
2970                 }
2971         }
2972
2973         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2974                 ifa = inet6_addr_lst[state->bucket];
2975                 goto try_again;
2976         }
2977
2978         return ifa;
2979 }
2980
2981 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2982 {
2983         struct inet6_ifaddr *ifa = if6_get_first(seq);
2984
2985         if (ifa)
2986                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2987                         --pos;
2988         return pos ? NULL : ifa;
2989 }
2990
2991 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2992         __acquires(addrconf_hash_lock)
2993 {
2994         read_lock_bh(&addrconf_hash_lock);
2995         return if6_get_idx(seq, *pos);
2996 }
2997
2998 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2999 {
3000         struct inet6_ifaddr *ifa;
3001
3002         ifa = if6_get_next(seq, v);
3003         ++*pos;
3004         return ifa;
3005 }
3006
3007 static void if6_seq_stop(struct seq_file *seq, void *v)
3008         __releases(addrconf_hash_lock)
3009 {
3010         read_unlock_bh(&addrconf_hash_lock);
3011 }
3012
3013 static int if6_seq_show(struct seq_file *seq, void *v)
3014 {
3015         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3016         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3017                    &ifp->addr,
3018                    ifp->idev->dev->ifindex,
3019                    ifp->prefix_len,
3020                    ifp->scope,
3021                    ifp->flags,
3022                    ifp->idev->dev->name);
3023         return 0;
3024 }
3025
3026 static const struct seq_operations if6_seq_ops = {
3027         .start  = if6_seq_start,
3028         .next   = if6_seq_next,
3029         .show   = if6_seq_show,
3030         .stop   = if6_seq_stop,
3031 };
3032
3033 static int if6_seq_open(struct inode *inode, struct file *file)
3034 {
3035         return seq_open_net(inode, file, &if6_seq_ops,
3036                             sizeof(struct if6_iter_state));
3037 }
3038
3039 static const struct file_operations if6_fops = {
3040         .owner          = THIS_MODULE,
3041         .open           = if6_seq_open,
3042         .read           = seq_read,
3043         .llseek         = seq_lseek,
3044         .release        = seq_release_net,
3045 };
3046
3047 static int __net_init if6_proc_net_init(struct net *net)
3048 {
3049         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3050                 return -ENOMEM;
3051         return 0;
3052 }
3053
3054 static void __net_exit if6_proc_net_exit(struct net *net)
3055 {
3056        proc_net_remove(net, "if_inet6");
3057 }
3058
3059 static struct pernet_operations if6_proc_net_ops = {
3060        .init = if6_proc_net_init,
3061        .exit = if6_proc_net_exit,
3062 };
3063
3064 int __init if6_proc_init(void)
3065 {
3066         return register_pernet_subsys(&if6_proc_net_ops);
3067 }
3068
3069 void if6_proc_exit(void)
3070 {
3071         unregister_pernet_subsys(&if6_proc_net_ops);
3072 }
3073 #endif  /* CONFIG_PROC_FS */
3074
3075 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3076 /* Check if address is a home address configured on any interface. */
3077 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3078 {
3079         int ret = 0;
3080         struct inet6_ifaddr * ifp;
3081         u8 hash = ipv6_addr_hash(addr);
3082         read_lock_bh(&addrconf_hash_lock);
3083         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3084                 if (!net_eq(dev_net(ifp->idev->dev), net))
3085                         continue;
3086                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3087                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3088                         ret = 1;
3089                         break;
3090                 }
3091         }
3092         read_unlock_bh(&addrconf_hash_lock);
3093         return ret;
3094 }
3095 #endif
3096
3097 /*
3098  *      Periodic address status verification
3099  */
3100
3101 static void addrconf_verify(unsigned long foo)
3102 {
3103         struct inet6_ifaddr *ifp;
3104         unsigned long now, next;
3105         int i;
3106
3107         spin_lock_bh(&addrconf_verify_lock);
3108         now = jiffies;
3109         next = now + ADDR_CHECK_FREQUENCY;
3110
3111         del_timer(&addr_chk_timer);
3112
3113         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3114
3115 restart:
3116                 read_lock(&addrconf_hash_lock);
3117                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3118                         unsigned long age;
3119 #ifdef CONFIG_IPV6_PRIVACY
3120                         unsigned long regen_advance;
3121 #endif
3122
3123                         if (ifp->flags & IFA_F_PERMANENT)
3124                                 continue;
3125
3126                         spin_lock(&ifp->lock);
3127                         age = (now - ifp->tstamp) / HZ;
3128
3129 #ifdef CONFIG_IPV6_PRIVACY
3130                         regen_advance = ifp->idev->cnf.regen_max_retry *
3131                                         ifp->idev->cnf.dad_transmits *
3132                                         ifp->idev->nd_parms->retrans_time / HZ;
3133 #endif
3134
3135                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3136                             age >= ifp->valid_lft) {
3137                                 spin_unlock(&ifp->lock);
3138                                 in6_ifa_hold(ifp);
3139                                 read_unlock(&addrconf_hash_lock);
3140                                 ipv6_del_addr(ifp);
3141                                 goto restart;
3142                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3143                                 spin_unlock(&ifp->lock);
3144                                 continue;
3145                         } else if (age >= ifp->prefered_lft) {
3146                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3147                                 int deprecate = 0;
3148
3149                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3150                                         deprecate = 1;
3151                                         ifp->flags |= IFA_F_DEPRECATED;
3152                                 }
3153
3154                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3155                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3156
3157                                 spin_unlock(&ifp->lock);
3158
3159                                 if (deprecate) {
3160                                         in6_ifa_hold(ifp);
3161                                         read_unlock(&addrconf_hash_lock);
3162
3163                                         ipv6_ifa_notify(0, ifp);
3164                                         in6_ifa_put(ifp);
3165                                         goto restart;
3166                                 }
3167 #ifdef CONFIG_IPV6_PRIVACY
3168                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3169                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3170                                 if (age >= ifp->prefered_lft - regen_advance) {
3171                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3172                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3173                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3174                                         if (!ifp->regen_count && ifpub) {
3175                                                 ifp->regen_count++;
3176                                                 in6_ifa_hold(ifp);
3177                                                 in6_ifa_hold(ifpub);
3178                                                 spin_unlock(&ifp->lock);
3179                                                 read_unlock(&addrconf_hash_lock);
3180                                                 spin_lock(&ifpub->lock);
3181                                                 ifpub->regen_count = 0;
3182                                                 spin_unlock(&ifpub->lock);
3183                                                 ipv6_create_tempaddr(ifpub, ifp);
3184                                                 in6_ifa_put(ifpub);
3185                                                 in6_ifa_put(ifp);
3186                                                 goto restart;
3187                                         }
3188                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3189                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3190                                 spin_unlock(&ifp->lock);
3191 #endif
3192                         } else {
3193                                 /* ifp->prefered_lft <= ifp->valid_lft */
3194                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3195                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3196                                 spin_unlock(&ifp->lock);
3197                         }
3198                 }
3199                 read_unlock(&addrconf_hash_lock);
3200         }
3201
3202         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3203         add_timer(&addr_chk_timer);
3204         spin_unlock_bh(&addrconf_verify_lock);
3205 }
3206
3207 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3208 {
3209         struct in6_addr *pfx = NULL;
3210
3211         if (addr)
3212                 pfx = nla_data(addr);
3213
3214         if (local) {
3215                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3216                         pfx = NULL;
3217                 else
3218                         pfx = nla_data(local);
3219         }
3220
3221         return pfx;
3222 }
3223
3224 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3225         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3226         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3227         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3228 };
3229
3230 static int
3231 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3232 {
3233         struct net *net = sock_net(skb->sk);
3234         struct ifaddrmsg *ifm;
3235         struct nlattr *tb[IFA_MAX+1];
3236         struct in6_addr *pfx;
3237         int err;
3238
3239         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3240         if (err < 0)
3241                 return err;
3242
3243         ifm = nlmsg_data(nlh);
3244         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3245         if (pfx == NULL)
3246                 return -EINVAL;
3247
3248         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3249 }
3250
3251 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3252                              u32 prefered_lft, u32 valid_lft)
3253 {
3254         u32 flags;
3255         clock_t expires;
3256         unsigned long timeout;
3257
3258         if (!valid_lft || (prefered_lft > valid_lft))
3259                 return -EINVAL;
3260
3261         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3262         if (addrconf_finite_timeout(timeout)) {
3263                 expires = jiffies_to_clock_t(timeout * HZ);
3264                 valid_lft = timeout;
3265                 flags = RTF_EXPIRES;
3266         } else {
3267                 expires = 0;
3268                 flags = 0;
3269                 ifa_flags |= IFA_F_PERMANENT;
3270         }
3271
3272         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3273         if (addrconf_finite_timeout(timeout)) {
3274                 if (timeout == 0)
3275                         ifa_flags |= IFA_F_DEPRECATED;
3276                 prefered_lft = timeout;
3277         }
3278
3279         spin_lock_bh(&ifp->lock);
3280         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3281         ifp->tstamp = jiffies;
3282         ifp->valid_lft = valid_lft;
3283         ifp->prefered_lft = prefered_lft;
3284
3285         spin_unlock_bh(&ifp->lock);
3286         if (!(ifp->flags&IFA_F_TENTATIVE))
3287                 ipv6_ifa_notify(0, ifp);
3288
3289         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3290                               expires, flags);
3291         addrconf_verify(0);
3292
3293         return 0;
3294 }
3295
3296 static int
3297 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3298 {
3299         struct net *net = sock_net(skb->sk);
3300         struct ifaddrmsg *ifm;
3301         struct nlattr *tb[IFA_MAX+1];
3302         struct in6_addr *pfx;
3303         struct inet6_ifaddr *ifa;
3304         struct net_device *dev;
3305         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3306         u8 ifa_flags;
3307         int err;
3308
3309         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3310         if (err < 0)
3311                 return err;
3312
3313         ifm = nlmsg_data(nlh);
3314         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3315         if (pfx == NULL)
3316                 return -EINVAL;
3317
3318         if (tb[IFA_CACHEINFO]) {
3319                 struct ifa_cacheinfo *ci;
3320
3321                 ci = nla_data(tb[IFA_CACHEINFO]);
3322                 valid_lft = ci->ifa_valid;
3323                 preferred_lft = ci->ifa_prefered;
3324         } else {
3325                 preferred_lft = INFINITY_LIFE_TIME;
3326                 valid_lft = INFINITY_LIFE_TIME;
3327         }
3328
3329         dev =  __dev_get_by_index(net, ifm->ifa_index);
3330         if (dev == NULL)
3331                 return -ENODEV;
3332
3333         /* We ignore other flags so far. */
3334         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3335
3336         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3337         if (ifa == NULL) {
3338                 /*
3339                  * It would be best to check for !NLM_F_CREATE here but
3340                  * userspace alreay relies on not having to provide this.
3341                  */
3342                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3343                                       ifm->ifa_prefixlen, ifa_flags,
3344                                       preferred_lft, valid_lft);
3345         }
3346
3347         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3348             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3349                 err = -EEXIST;
3350         else
3351                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3352
3353         in6_ifa_put(ifa);
3354
3355         return err;
3356 }
3357
3358 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3359                           u8 scope, int ifindex)
3360 {
3361         struct ifaddrmsg *ifm;
3362
3363         ifm = nlmsg_data(nlh);
3364         ifm->ifa_family = AF_INET6;
3365         ifm->ifa_prefixlen = prefixlen;
3366         ifm->ifa_flags = flags;
3367         ifm->ifa_scope = scope;
3368         ifm->ifa_index = ifindex;
3369 }
3370
3371 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3372                          unsigned long tstamp, u32 preferred, u32 valid)
3373 {
3374         struct ifa_cacheinfo ci;
3375
3376         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3377                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3378         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3379                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3380         ci.ifa_prefered = preferred;
3381         ci.ifa_valid = valid;
3382
3383         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3384 }
3385
3386 static inline int rt_scope(int ifa_scope)
3387 {
3388         if (ifa_scope & IFA_HOST)
3389                 return RT_SCOPE_HOST;
3390         else if (ifa_scope & IFA_LINK)
3391                 return RT_SCOPE_LINK;
3392         else if (ifa_scope & IFA_SITE)
3393                 return RT_SCOPE_SITE;
3394         else
3395                 return RT_SCOPE_UNIVERSE;
3396 }
3397
3398 static inline int inet6_ifaddr_msgsize(void)
3399 {
3400         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3401                + nla_total_size(16) /* IFA_ADDRESS */
3402                + nla_total_size(sizeof(struct ifa_cacheinfo));
3403 }
3404
3405 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3406                              u32 pid, u32 seq, int event, unsigned int flags)
3407 {
3408         struct nlmsghdr  *nlh;
3409         u32 preferred, valid;
3410
3411         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3412         if (nlh == NULL)
3413                 return -EMSGSIZE;
3414
3415         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3416                       ifa->idev->dev->ifindex);
3417
3418         if (!(ifa->flags&IFA_F_PERMANENT)) {
3419                 preferred = ifa->prefered_lft;
3420                 valid = ifa->valid_lft;
3421                 if (preferred != INFINITY_LIFE_TIME) {
3422                         long tval = (jiffies - ifa->tstamp)/HZ;
3423                         if (preferred > tval)
3424                                 preferred -= tval;
3425                         else
3426                                 preferred = 0;
3427                         if (valid != INFINITY_LIFE_TIME)
3428                                 valid -= tval;
3429                 }
3430         } else {
3431                 preferred = INFINITY_LIFE_TIME;
3432                 valid = INFINITY_LIFE_TIME;
3433         }
3434
3435         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3436             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3437                 nlmsg_cancel(skb, nlh);
3438                 return -EMSGSIZE;
3439         }
3440
3441         return nlmsg_end(skb, nlh);
3442 }
3443
3444 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3445                                 u32 pid, u32 seq, int event, u16 flags)
3446 {
3447         struct nlmsghdr  *nlh;
3448         u8 scope = RT_SCOPE_UNIVERSE;
3449         int ifindex = ifmca->idev->dev->ifindex;
3450
3451         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3452                 scope = RT_SCOPE_SITE;
3453
3454         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3455         if (nlh == NULL)
3456                 return -EMSGSIZE;
3457
3458         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3459         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3460             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3461                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3462                 nlmsg_cancel(skb, nlh);
3463                 return -EMSGSIZE;
3464         }
3465
3466         return nlmsg_end(skb, nlh);
3467 }
3468
3469 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3470                                 u32 pid, u32 seq, int event, unsigned int flags)
3471 {
3472         struct nlmsghdr  *nlh;
3473         u8 scope = RT_SCOPE_UNIVERSE;
3474         int ifindex = ifaca->aca_idev->dev->ifindex;
3475
3476         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3477                 scope = RT_SCOPE_SITE;
3478
3479         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3480         if (nlh == NULL)
3481                 return -EMSGSIZE;
3482
3483         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3484         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3485             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3486                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3487                 nlmsg_cancel(skb, nlh);
3488                 return -EMSGSIZE;
3489         }
3490
3491         return nlmsg_end(skb, nlh);
3492 }
3493
3494 enum addr_type_t
3495 {
3496         UNICAST_ADDR,
3497         MULTICAST_ADDR,
3498         ANYCAST_ADDR,
3499 };
3500
3501 /* called with rcu_read_lock() */
3502 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3503                           struct netlink_callback *cb, enum addr_type_t type,
3504                           int s_ip_idx, int *p_ip_idx)
3505 {
3506         struct inet6_ifaddr *ifa;
3507         struct ifmcaddr6 *ifmca;
3508         struct ifacaddr6 *ifaca;
3509         int err = 1;
3510         int ip_idx = *p_ip_idx;
3511
3512         read_lock_bh(&idev->lock);
3513         switch (type) {
3514         case UNICAST_ADDR:
3515                 /* unicast address incl. temp addr */
3516                 for (ifa = idev->addr_list; ifa;
3517                      ifa = ifa->if_next, ip_idx++) {
3518                         if (ip_idx < s_ip_idx)
3519                                 continue;
3520                         err = inet6_fill_ifaddr(skb, ifa,
3521                                                 NETLINK_CB(cb->skb).pid,
3522                                                 cb->nlh->nlmsg_seq,
3523                                                 RTM_NEWADDR,
3524                                                 NLM_F_MULTI);
3525                         if (err <= 0)
3526                                 break;
3527                 }
3528                 break;
3529         case MULTICAST_ADDR:
3530                 /* multicast address */
3531                 for (ifmca = idev->mc_list; ifmca;
3532                      ifmca = ifmca->next, ip_idx++) {
3533                         if (ip_idx < s_ip_idx)
3534                                 continue;
3535                         err = inet6_fill_ifmcaddr(skb, ifmca,
3536                                                   NETLINK_CB(cb->skb).pid,
3537                                                   cb->nlh->nlmsg_seq,
3538                                                   RTM_GETMULTICAST,
3539                                                   NLM_F_MULTI);
3540                         if (err <= 0)
3541                                 break;
3542                 }
3543                 break;
3544         case ANYCAST_ADDR:
3545                 /* anycast address */
3546                 for (ifaca = idev->ac_list; ifaca;
3547                      ifaca = ifaca->aca_next, ip_idx++) {
3548                         if (ip_idx < s_ip_idx)
3549                                 continue;
3550                         err = inet6_fill_ifacaddr(skb, ifaca,
3551                                                   NETLINK_CB(cb->skb).pid,
3552                                                   cb->nlh->nlmsg_seq,
3553                                                   RTM_GETANYCAST,
3554                                                   NLM_F_MULTI);
3555                         if (err <= 0)
3556                                 break;
3557                 }
3558                 break;
3559         default:
3560                 break;
3561         }
3562         read_unlock_bh(&idev->lock);
3563         *p_ip_idx = ip_idx;
3564         return err;
3565 }
3566
3567 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3568                            enum addr_type_t type)
3569 {
3570         struct net *net = sock_net(skb->sk);
3571         int h, s_h;
3572         int idx, ip_idx;
3573         int s_idx, s_ip_idx;
3574         struct net_device *dev;
3575         struct inet6_dev *idev;
3576         struct hlist_head *head;
3577         struct hlist_node *node;
3578
3579         s_h = cb->args[0];
3580         s_idx = idx = cb->args[1];
3581         s_ip_idx = ip_idx = cb->args[2];
3582
3583         rcu_read_lock();
3584         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3585                 idx = 0;
3586                 head = &net->dev_index_head[h];
3587                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3588                         if (idx < s_idx)
3589                                 goto cont;
3590                         if (idx > s_idx)
3591                                 s_ip_idx = 0;
3592                         ip_idx = 0;
3593                         if ((idev = __in6_dev_get(dev)) == NULL)
3594                                 goto cont;
3595
3596                         if (in6_dump_addrs(idev, skb, cb, type,
3597                                            s_ip_idx, &ip_idx) <= 0)
3598                                 goto done;
3599 cont:
3600                         idx++;
3601                 }
3602         }
3603 done:
3604         rcu_read_unlock();
3605         cb->args[0] = h;
3606         cb->args[1] = idx;
3607         cb->args[2] = ip_idx;
3608
3609         return skb->len;
3610 }
3611
3612 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3613 {
3614         enum addr_type_t type = UNICAST_ADDR;
3615
3616         return inet6_dump_addr(skb, cb, type);
3617 }
3618
3619 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3620 {
3621         enum addr_type_t type = MULTICAST_ADDR;
3622
3623         return inet6_dump_addr(skb, cb, type);
3624 }
3625
3626
3627 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3628 {
3629         enum addr_type_t type = ANYCAST_ADDR;
3630
3631         return inet6_dump_addr(skb, cb, type);
3632 }
3633
3634 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3635                              void *arg)
3636 {
3637         struct net *net = sock_net(in_skb->sk);
3638         struct ifaddrmsg *ifm;
3639         struct nlattr *tb[IFA_MAX+1];
3640         struct in6_addr *addr = NULL;
3641         struct net_device *dev = NULL;
3642         struct inet6_ifaddr *ifa;
3643         struct sk_buff *skb;
3644         int err;
3645
3646         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3647         if (err < 0)
3648                 goto errout;
3649
3650         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3651         if (addr == NULL) {
3652                 err = -EINVAL;
3653                 goto errout;
3654         }
3655
3656         ifm = nlmsg_data(nlh);
3657         if (ifm->ifa_index)
3658                 dev = __dev_get_by_index(net, ifm->ifa_index);
3659
3660         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3661                 err = -EADDRNOTAVAIL;
3662                 goto errout;
3663         }
3664
3665         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3666                 err = -ENOBUFS;
3667                 goto errout_ifa;
3668         }
3669
3670         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3671                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3672         if (err < 0) {
3673                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3674                 WARN_ON(err == -EMSGSIZE);
3675                 kfree_skb(skb);
3676                 goto errout_ifa;
3677         }
3678         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3679 errout_ifa:
3680         in6_ifa_put(ifa);
3681 errout:
3682         return err;
3683 }
3684
3685 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3686 {
3687         struct sk_buff *skb;
3688         struct net *net = dev_net(ifa->idev->dev);
3689         int err = -ENOBUFS;
3690
3691         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3692         if (skb == NULL)
3693                 goto errout;
3694
3695         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3696         if (err < 0) {
3697                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3698                 WARN_ON(err == -EMSGSIZE);
3699                 kfree_skb(skb);
3700                 goto errout;
3701         }
3702         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3703         return;
3704 errout:
3705         if (err < 0)
3706                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3707 }
3708
3709 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3710                                 __s32 *array, int bytes)
3711 {
3712         BUG_ON(bytes < (DEVCONF_MAX * 4));
3713
3714         memset(array, 0, bytes);
3715         array[DEVCONF_FORWARDING] = cnf->forwarding;
3716         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3717         array[DEVCONF_MTU6] = cnf->mtu6;
3718         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3719         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3720         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3721         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3722         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3723         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3724         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3725         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3726 #ifdef CONFIG_IPV6_PRIVACY
3727         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3728         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3729         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3730         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3731         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3732 #endif
3733         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3734         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3735         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3736 #ifdef CONFIG_IPV6_ROUTER_PREF
3737         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3738         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3739 #ifdef CONFIG_IPV6_ROUTE_INFO
3740         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3741 #endif
3742 #endif
3743         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3744         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3745 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3746         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3747 #endif
3748 #ifdef CONFIG_IPV6_MROUTE
3749         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3750 #endif
3751         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3752         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3753         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3754 }
3755
3756 static inline size_t inet6_if_nlmsg_size(void)
3757 {
3758         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3759                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3760                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3761                + nla_total_size(4) /* IFLA_MTU */
3762                + nla_total_size(4) /* IFLA_LINK */
3763                + nla_total_size( /* IFLA_PROTINFO */
3764                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3765                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3766                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3767                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3768                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3769                  );
3770 }
3771
3772 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3773                                       int items, int bytes)
3774 {
3775         int i;
3776         int pad = bytes - sizeof(u64) * items;
3777         BUG_ON(pad < 0);
3778
3779         /* Use put_unaligned() because stats may not be aligned for u64. */
3780         put_unaligned(items, &stats[0]);
3781         for (i = 1; i < items; i++)
3782                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3783
3784         memset(&stats[items], 0, pad);
3785 }
3786
3787 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3788                              int bytes)
3789 {
3790         switch(attrtype) {
3791         case IFLA_INET6_STATS:
3792                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3793                 break;
3794         case IFLA_INET6_ICMP6STATS:
3795                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3796                 break;
3797         }
3798 }
3799
3800 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3801                              u32 pid, u32 seq, int event, unsigned int flags)
3802 {
3803         struct net_device *dev = idev->dev;
3804         struct nlattr *nla;
3805         struct ifinfomsg *hdr;
3806         struct nlmsghdr *nlh;
3807         void *protoinfo;
3808         struct ifla_cacheinfo ci;
3809
3810         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3811         if (nlh == NULL)
3812                 return -EMSGSIZE;
3813
3814         hdr = nlmsg_data(nlh);
3815         hdr->ifi_family = AF_INET6;
3816         hdr->__ifi_pad = 0;
3817         hdr->ifi_type = dev->type;
3818         hdr->ifi_index = dev->ifindex;
3819         hdr->ifi_flags = dev_get_flags(dev);
3820         hdr->ifi_change = 0;
3821
3822         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3823
3824         if (dev->addr_len)
3825                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3826
3827         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3828         if (dev->ifindex != dev->iflink)
3829                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3830
3831         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3832         if (protoinfo == NULL)
3833                 goto nla_put_failure;
3834
3835         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3836
3837         ci.max_reasm_len = IPV6_MAXPLEN;
3838         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3839                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3840         ci.reachable_time = idev->nd_parms->reachable_time;
3841         ci.retrans_time = idev->nd_parms->retrans_time;
3842         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3843
3844         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3845         if (nla == NULL)
3846                 goto nla_put_failure;
3847         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3848
3849         /* XXX - MC not implemented */
3850
3851         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3852         if (nla == NULL)
3853                 goto nla_put_failure;
3854         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3855
3856         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3857         if (nla == NULL)
3858                 goto nla_put_failure;
3859         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3860
3861         nla_nest_end(skb, protoinfo);
3862         return nlmsg_end(skb, nlh);
3863
3864 nla_put_failure:
3865         nlmsg_cancel(skb, nlh);
3866         return -EMSGSIZE;
3867 }
3868
3869 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3870 {
3871         struct net *net = sock_net(skb->sk);
3872         int h, s_h;
3873         int idx = 0, s_idx;
3874         struct net_device *dev;
3875         struct inet6_dev *idev;
3876         struct hlist_head *head;
3877         struct hlist_node *node;
3878
3879         s_h = cb->args[0];
3880         s_idx = cb->args[1];
3881
3882         rcu_read_lock();
3883         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3884                 idx = 0;
3885                 head = &net->dev_index_head[h];
3886                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3887                         if (idx < s_idx)
3888                                 goto cont;
3889                         idev = __in6_dev_get(dev);
3890                         if (!idev)
3891                                 goto cont;
3892                         if (inet6_fill_ifinfo(skb, idev,
3893                                               NETLINK_CB(cb->skb).pid,
3894                                               cb->nlh->nlmsg_seq,
3895                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3896                                 goto out;
3897 cont:
3898                         idx++;
3899                 }
3900         }
3901 out:
3902         rcu_read_unlock();
3903         cb->args[1] = idx;
3904         cb->args[0] = h;
3905
3906         return skb->len;
3907 }
3908
3909 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3910 {
3911         struct sk_buff *skb;
3912         struct net *net = dev_net(idev->dev);
3913         int err = -ENOBUFS;
3914
3915         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3916         if (skb == NULL)
3917                 goto errout;
3918
3919         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3920         if (err < 0) {
3921                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3922                 WARN_ON(err == -EMSGSIZE);
3923                 kfree_skb(skb);
3924                 goto errout;
3925         }
3926         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3927         return;
3928 errout:
3929         if (err < 0)
3930                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3931 }
3932
3933 static inline size_t inet6_prefix_nlmsg_size(void)
3934 {
3935         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3936                + nla_total_size(sizeof(struct in6_addr))
3937                + nla_total_size(sizeof(struct prefix_cacheinfo));
3938 }
3939
3940 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3941                              struct prefix_info *pinfo, u32 pid, u32 seq,
3942                              int event, unsigned int flags)
3943 {
3944         struct prefixmsg *pmsg;
3945         struct nlmsghdr *nlh;
3946         struct prefix_cacheinfo ci;
3947
3948         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3949         if (nlh == NULL)
3950                 return -EMSGSIZE;
3951
3952         pmsg = nlmsg_data(nlh);
3953         pmsg->prefix_family = AF_INET6;
3954         pmsg->prefix_pad1 = 0;
3955         pmsg->prefix_pad2 = 0;
3956         pmsg->prefix_ifindex = idev->dev->ifindex;
3957         pmsg->prefix_len = pinfo->prefix_len;
3958         pmsg->prefix_type = pinfo->type;
3959         pmsg->prefix_pad3 = 0;
3960         pmsg->prefix_flags = 0;
3961         if (pinfo->onlink)
3962                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3963         if (pinfo->autoconf)
3964                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3965
3966         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3967
3968         ci.preferred_time = ntohl(pinfo->prefered);
3969         ci.valid_time = ntohl(pinfo->valid);
3970         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3971
3972         return nlmsg_end(skb, nlh);
3973
3974 nla_put_failure:
3975         nlmsg_cancel(skb, nlh);
3976         return -EMSGSIZE;
3977 }
3978
3979 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3980                          struct prefix_info *pinfo)
3981 {
3982         struct sk_buff *skb;
3983         struct net *net = dev_net(idev->dev);
3984         int err = -ENOBUFS;
3985
3986         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3987         if (skb == NULL)
3988                 goto errout;
3989
3990         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3991         if (err < 0) {
3992                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3993                 WARN_ON(err == -EMSGSIZE);
3994                 kfree_skb(skb);
3995                 goto errout;
3996         }
3997         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3998         return;
3999 errout:
4000         if (err < 0)
4001                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4002 }
4003
4004 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4005 {
4006         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4007
4008         switch (event) {
4009         case RTM_NEWADDR:
4010                 /*
4011                  * If the address was optimistic
4012                  * we inserted the route at the start of
4013                  * our DAD process, so we don't need
4014                  * to do it again
4015                  */
4016                 if (!(ifp->rt->rt6i_node))
4017                         ip6_ins_rt(ifp->rt);
4018                 if (ifp->idev->cnf.forwarding)
4019                         addrconf_join_anycast(ifp);
4020                 break;
4021         case RTM_DELADDR:
4022                 if (ifp->idev->cnf.forwarding)
4023                         addrconf_leave_anycast(ifp);
4024                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4025                 dst_hold(&ifp->rt->u.dst);
4026                 if (ip6_del_rt(ifp->rt))
4027                         dst_free(&ifp->rt->u.dst);
4028                 break;
4029         }
4030 }
4031
4032 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4033 {
4034         rcu_read_lock_bh();
4035         if (likely(ifp->idev->dead == 0))
4036                 __ipv6_ifa_notify(event, ifp);
4037         rcu_read_unlock_bh();
4038 }
4039
4040 #ifdef CONFIG_SYSCTL
4041
4042 static
4043 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4044                            void __user *buffer, size_t *lenp, loff_t *ppos)
4045 {
4046         int *valp = ctl->data;
4047         int val = *valp;
4048         loff_t pos = *ppos;
4049         int ret;
4050
4051         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4052
4053         if (write)
4054                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4055         if (ret)
4056                 *ppos = pos;
4057         return ret;
4058 }
4059
4060 static void dev_disable_change(struct inet6_dev *idev)
4061 {
4062         if (!idev || !idev->dev)
4063                 return;
4064
4065         if (idev->cnf.disable_ipv6)
4066                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4067         else
4068                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4069 }
4070
4071 static void addrconf_disable_change(struct net *net, __s32 newf)
4072 {
4073         struct net_device *dev;
4074         struct inet6_dev *idev;
4075
4076         rcu_read_lock();
4077         for_each_netdev_rcu(net, dev) {
4078                 idev = __in6_dev_get(dev);
4079                 if (idev) {
4080                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4081                         idev->cnf.disable_ipv6 = newf;
4082                         if (changed)
4083                                 dev_disable_change(idev);
4084                 }
4085         }
4086         rcu_read_unlock();
4087 }
4088
4089 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4090 {
4091         struct net *net;
4092
4093         net = (struct net *)table->extra2;
4094
4095         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4096                 return 0;
4097
4098         if (!rtnl_trylock()) {
4099                 /* Restore the original values before restarting */
4100                 *p = old;
4101                 return restart_syscall();
4102         }
4103
4104         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4105                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4106                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4107                 addrconf_disable_change(net, newf);
4108         } else if ((!*p) ^ (!old))
4109                 dev_disable_change((struct inet6_dev *)table->extra1);
4110
4111         rtnl_unlock();
4112         return 0;
4113 }
4114
4115 static
4116 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4117                             void __user *buffer, size_t *lenp, loff_t *ppos)
4118 {
4119         int *valp = ctl->data;
4120         int val = *valp;
4121         loff_t pos = *ppos;
4122         int ret;
4123
4124         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4125
4126         if (write)
4127                 ret = addrconf_disable_ipv6(ctl, valp, val);
4128         if (ret)
4129                 *ppos = pos;
4130         return ret;
4131 }
4132
4133 static struct addrconf_sysctl_table
4134 {
4135         struct ctl_table_header *sysctl_header;
4136         ctl_table addrconf_vars[DEVCONF_MAX+1];
4137         char *dev_name;
4138 } addrconf_sysctl __read_mostly = {
4139         .sysctl_header = NULL,
4140         .addrconf_vars = {
4141                 {
4142                         .procname       =       "forwarding",
4143                         .data           =       &ipv6_devconf.forwarding,
4144                         .maxlen         =       sizeof(int),
4145                         .mode           =       0644,
4146                         .proc_handler   =       addrconf_sysctl_forward,
4147                 },
4148                 {
4149                         .procname       =       "hop_limit",
4150                         .data           =       &ipv6_devconf.hop_limit,
4151                         .maxlen         =       sizeof(int),
4152                         .mode           =       0644,
4153                         .proc_handler   =       proc_dointvec,
4154                 },
4155                 {
4156                         .procname       =       "mtu",
4157                         .data           =       &ipv6_devconf.mtu6,
4158                         .maxlen         =       sizeof(int),
4159                         .mode           =       0644,
4160                         .proc_handler   =       proc_dointvec,
4161                 },
4162                 {
4163                         .procname       =       "accept_ra",
4164                         .data           =       &ipv6_devconf.accept_ra,
4165                         .maxlen         =       sizeof(int),
4166                         .mode           =       0644,
4167                         .proc_handler   =       proc_dointvec,
4168                 },
4169                 {
4170                         .procname       =       "accept_redirects",
4171                         .data           =       &ipv6_devconf.accept_redirects,
4172                         .maxlen         =       sizeof(int),
4173                         .mode           =       0644,
4174                         .proc_handler   =       proc_dointvec,
4175                 },
4176                 {
4177                         .procname       =       "autoconf",
4178                         .data           =       &ipv6_devconf.autoconf,
4179                         .maxlen         =       sizeof(int),
4180                         .mode           =       0644,
4181                         .proc_handler   =       proc_dointvec,
4182                 },
4183                 {
4184                         .procname       =       "dad_transmits",
4185                         .data           =       &ipv6_devconf.dad_transmits,
4186                         .maxlen         =       sizeof(int),
4187                         .mode           =       0644,
4188                         .proc_handler   =       proc_dointvec,
4189                 },
4190                 {
4191                         .procname       =       "router_solicitations",
4192                         .data           =       &ipv6_devconf.rtr_solicits,
4193                         .maxlen         =       sizeof(int),
4194                         .mode           =       0644,
4195                         .proc_handler   =       proc_dointvec,
4196                 },
4197                 {
4198                         .procname       =       "router_solicitation_interval",
4199                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4200                         .maxlen         =       sizeof(int),
4201                         .mode           =       0644,
4202                         .proc_handler   =       proc_dointvec_jiffies,
4203                 },
4204                 {
4205                         .procname       =       "router_solicitation_delay",
4206                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4207                         .maxlen         =       sizeof(int),
4208                         .mode           =       0644,
4209                         .proc_handler   =       proc_dointvec_jiffies,
4210                 },
4211                 {
4212                         .procname       =       "force_mld_version",
4213                         .data           =       &ipv6_devconf.force_mld_version,
4214                         .maxlen         =       sizeof(int),
4215                         .mode           =       0644,
4216                         .proc_handler   =       proc_dointvec,
4217                 },
4218 #ifdef CONFIG_IPV6_PRIVACY
4219                 {
4220                         .procname       =       "use_tempaddr",
4221                         .data           =       &ipv6_devconf.use_tempaddr,
4222                         .maxlen         =       sizeof(int),
4223                         .mode           =       0644,
4224                         .proc_handler   =       proc_dointvec,
4225                 },
4226                 {
4227                         .procname       =       "temp_valid_lft",
4228                         .data           =       &ipv6_devconf.temp_valid_lft,
4229                         .maxlen         =       sizeof(int),
4230                         .mode           =       0644,
4231                         .proc_handler   =       proc_dointvec,
4232                 },
4233                 {
4234                         .procname       =       "temp_prefered_lft",
4235                         .data           =       &ipv6_devconf.temp_prefered_lft,
4236                         .maxlen         =       sizeof(int),
4237                         .mode           =       0644,
4238                         .proc_handler   =       proc_dointvec,
4239                 },
4240                 {
4241                         .procname       =       "regen_max_retry",
4242                         .data           =       &ipv6_devconf.regen_max_retry,
4243                         .maxlen         =       sizeof(int),
4244                         .mode           =       0644,
4245                         .proc_handler   =       proc_dointvec,
4246                 },
4247                 {
4248                         .procname       =       "max_desync_factor",
4249                         .data           =       &ipv6_devconf.max_desync_factor,
4250                         .maxlen         =       sizeof(int),
4251                         .mode           =       0644,
4252                         .proc_handler   =       proc_dointvec,
4253                 },
4254 #endif
4255                 {
4256                         .procname       =       "max_addresses",
4257                         .data           =       &ipv6_devconf.max_addresses,
4258                         .maxlen         =       sizeof(int),
4259                         .mode           =       0644,
4260                         .proc_handler   =       proc_dointvec,
4261                 },
4262                 {
4263                         .procname       =       "accept_ra_defrtr",
4264                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4265                         .maxlen         =       sizeof(int),
4266                         .mode           =       0644,
4267                         .proc_handler   =       proc_dointvec,
4268                 },
4269                 {
4270                         .procname       =       "accept_ra_pinfo",
4271                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4272                         .maxlen         =       sizeof(int),
4273                         .mode           =       0644,
4274                         .proc_handler   =       proc_dointvec,
4275                 },
4276 #ifdef CONFIG_IPV6_ROUTER_PREF
4277                 {
4278                         .procname       =       "accept_ra_rtr_pref",
4279                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4280                         .maxlen         =       sizeof(int),
4281                         .mode           =       0644,
4282                         .proc_handler   =       proc_dointvec,
4283                 },
4284                 {
4285                         .procname       =       "router_probe_interval",
4286                         .data           =       &ipv6_devconf.rtr_probe_interval,
4287                         .maxlen         =       sizeof(int),
4288                         .mode           =       0644,
4289                         .proc_handler   =       proc_dointvec_jiffies,
4290                 },
4291 #ifdef CONFIG_IPV6_ROUTE_INFO
4292                 {
4293                         .procname       =       "accept_ra_rt_info_max_plen",
4294                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4295                         .maxlen         =       sizeof(int),
4296                         .mode           =       0644,
4297                         .proc_handler   =       proc_dointvec,
4298                 },
4299 #endif
4300 #endif
4301                 {
4302                         .procname       =       "proxy_ndp",
4303                         .data           =       &ipv6_devconf.proxy_ndp,
4304                         .maxlen         =       sizeof(int),
4305                         .mode           =       0644,
4306                         .proc_handler   =       proc_dointvec,
4307                 },
4308                 {
4309                         .procname       =       "accept_source_route",
4310                         .data           =       &ipv6_devconf.accept_source_route,
4311                         .maxlen         =       sizeof(int),
4312                         .mode           =       0644,
4313                         .proc_handler   =       proc_dointvec,
4314                 },
4315 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4316                 {
4317                         .procname       =       "optimistic_dad",
4318                         .data           =       &ipv6_devconf.optimistic_dad,
4319                         .maxlen         =       sizeof(int),
4320                         .mode           =       0644,
4321                         .proc_handler   =       proc_dointvec,
4322
4323                 },
4324 #endif
4325 #ifdef CONFIG_IPV6_MROUTE
4326                 {
4327                         .procname       =       "mc_forwarding",
4328                         .data           =       &ipv6_devconf.mc_forwarding,
4329                         .maxlen         =       sizeof(int),
4330                         .mode           =       0444,
4331                         .proc_handler   =       proc_dointvec,
4332                 },
4333 #endif
4334                 {
4335                         .procname       =       "disable_ipv6",
4336                         .data           =       &ipv6_devconf.disable_ipv6,
4337                         .maxlen         =       sizeof(int),
4338                         .mode           =       0644,
4339                         .proc_handler   =       addrconf_sysctl_disable,
4340                 },
4341                 {
4342                         .procname       =       "accept_dad",
4343                         .data           =       &ipv6_devconf.accept_dad,
4344                         .maxlen         =       sizeof(int),
4345                         .mode           =       0644,
4346                         .proc_handler   =       proc_dointvec,
4347                 },
4348                 {
4349                         .procname       = "force_tllao",
4350                         .data           = &ipv6_devconf.force_tllao,
4351                         .maxlen         = sizeof(int),
4352                         .mode           = 0644,
4353                         .proc_handler   = proc_dointvec
4354                 },
4355                 {
4356                         /* sentinel */
4357                 }
4358         },
4359 };
4360
4361 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4362                 struct inet6_dev *idev, struct ipv6_devconf *p)
4363 {
4364         int i;
4365         struct addrconf_sysctl_table *t;
4366
4367 #define ADDRCONF_CTL_PATH_DEV   3
4368
4369         struct ctl_path addrconf_ctl_path[] = {
4370                 { .procname = "net", },
4371                 { .procname = "ipv6", },
4372                 { .procname = "conf", },
4373                 { /* to be set */ },
4374                 { },
4375         };
4376
4377
4378         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4379         if (t == NULL)
4380                 goto out;
4381
4382         for (i=0; t->addrconf_vars[i].data; i++) {
4383                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4384                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4385                 t->addrconf_vars[i].extra2 = net;
4386         }
4387
4388         /*
4389          * Make a copy of dev_name, because '.procname' is regarded as const
4390          * by sysctl and we wouldn't want anyone to change it under our feet
4391          * (see SIOCSIFNAME).
4392          */
4393         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4394         if (!t->dev_name)
4395                 goto free;
4396
4397         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4398
4399         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4400                         t->addrconf_vars);
4401         if (t->sysctl_header == NULL)
4402                 goto free_procname;
4403
4404         p->sysctl = t;
4405         return 0;
4406
4407 free_procname:
4408         kfree(t->dev_name);
4409 free:
4410         kfree(t);
4411 out:
4412         return -ENOBUFS;
4413 }
4414
4415 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4416 {
4417         struct addrconf_sysctl_table *t;
4418
4419         if (p->sysctl == NULL)
4420                 return;
4421
4422         t = p->sysctl;
4423         p->sysctl = NULL;
4424         unregister_sysctl_table(t->sysctl_header);
4425         kfree(t->dev_name);
4426         kfree(t);
4427 }
4428
4429 static void addrconf_sysctl_register(struct inet6_dev *idev)
4430 {
4431         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4432                               &ndisc_ifinfo_sysctl_change);
4433         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4434                                         idev, &idev->cnf);
4435 }
4436
4437 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4438 {
4439         __addrconf_sysctl_unregister(&idev->cnf);
4440         neigh_sysctl_unregister(idev->nd_parms);
4441 }
4442
4443
4444 #endif
4445
4446 static int __net_init addrconf_init_net(struct net *net)
4447 {
4448         int err;
4449         struct ipv6_devconf *all, *dflt;
4450
4451         err = -ENOMEM;
4452         all = &ipv6_devconf;
4453         dflt = &ipv6_devconf_dflt;
4454
4455         if (!net_eq(net, &init_net)) {
4456                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4457                 if (all == NULL)
4458                         goto err_alloc_all;
4459
4460                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4461                 if (dflt == NULL)
4462                         goto err_alloc_dflt;
4463         } else {
4464                 /* these will be inherited by all namespaces */
4465                 dflt->autoconf = ipv6_defaults.autoconf;
4466                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4467         }
4468
4469         net->ipv6.devconf_all = all;
4470         net->ipv6.devconf_dflt = dflt;
4471
4472 #ifdef CONFIG_SYSCTL
4473         err = __addrconf_sysctl_register(net, "all", NULL, all);
4474         if (err < 0)
4475                 goto err_reg_all;
4476
4477         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4478         if (err < 0)
4479                 goto err_reg_dflt;
4480 #endif
4481         return 0;
4482
4483 #ifdef CONFIG_SYSCTL
4484 err_reg_dflt:
4485         __addrconf_sysctl_unregister(all);
4486 err_reg_all:
4487         kfree(dflt);
4488 #endif
4489 err_alloc_dflt:
4490         kfree(all);
4491 err_alloc_all:
4492         return err;
4493 }
4494
4495 static void __net_exit addrconf_exit_net(struct net *net)
4496 {
4497 #ifdef CONFIG_SYSCTL
4498         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4499         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4500 #endif
4501         if (!net_eq(net, &init_net)) {
4502                 kfree(net->ipv6.devconf_dflt);
4503                 kfree(net->ipv6.devconf_all);
4504         }
4505 }
4506
4507 static struct pernet_operations addrconf_ops = {
4508         .init = addrconf_init_net,
4509         .exit = addrconf_exit_net,
4510 };
4511
4512 /*
4513  *      Device notifier
4514  */
4515
4516 int register_inet6addr_notifier(struct notifier_block *nb)
4517 {
4518         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4519 }
4520
4521 EXPORT_SYMBOL(register_inet6addr_notifier);
4522
4523 int unregister_inet6addr_notifier(struct notifier_block *nb)
4524 {
4525         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4526 }
4527
4528 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4529
4530 /*
4531  *      Init / cleanup code
4532  */
4533
4534 int __init addrconf_init(void)
4535 {
4536         int err;
4537
4538         if ((err = ipv6_addr_label_init()) < 0) {
4539                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4540                         err);
4541                 return err;
4542         }
4543
4544         register_pernet_subsys(&addrconf_ops);
4545
4546         /* The addrconf netdev notifier requires that loopback_dev
4547          * has it's ipv6 private information allocated and setup
4548          * before it can bring up and give link-local addresses
4549          * to other devices which are up.
4550          *
4551          * Unfortunately, loopback_dev is not necessarily the first
4552          * entry in the global dev_base list of net devices.  In fact,
4553          * it is likely to be the very last entry on that list.
4554          * So this causes the notifier registry below to try and
4555          * give link-local addresses to all devices besides loopback_dev
4556          * first, then loopback_dev, which cases all the non-loopback_dev
4557          * devices to fail to get a link-local address.
4558          *
4559          * So, as a temporary fix, allocate the ipv6 structure for
4560          * loopback_dev first by hand.
4561          * Longer term, all of the dependencies ipv6 has upon the loopback
4562          * device and it being up should be removed.
4563          */
4564         rtnl_lock();
4565         if (!ipv6_add_dev(init_net.loopback_dev))
4566                 err = -ENOMEM;
4567         rtnl_unlock();
4568         if (err)
4569                 goto errlo;
4570
4571         register_netdevice_notifier(&ipv6_dev_notf);
4572
4573         addrconf_verify(0);
4574
4575         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4576         if (err < 0)
4577                 goto errout;
4578
4579         /* Only the first call to __rtnl_register can fail */
4580         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4581         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4582         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4583         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4584         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4585
4586         ipv6_addr_label_rtnl_register();
4587
4588         return 0;
4589 errout:
4590         unregister_netdevice_notifier(&ipv6_dev_notf);
4591 errlo:
4592         unregister_pernet_subsys(&addrconf_ops);
4593
4594         return err;
4595 }
4596
4597 void addrconf_cleanup(void)
4598 {
4599         struct inet6_ifaddr *ifa;
4600         struct net_device *dev;
4601         int i;
4602
4603         unregister_netdevice_notifier(&ipv6_dev_notf);
4604         unregister_pernet_subsys(&addrconf_ops);
4605
4606         rtnl_lock();
4607
4608         /* clean dev list */
4609         for_each_netdev(&init_net, dev) {
4610                 if (__in6_dev_get(dev) == NULL)
4611                         continue;
4612                 addrconf_ifdown(dev, 1);
4613         }
4614         addrconf_ifdown(init_net.loopback_dev, 2);
4615
4616         /*
4617          *      Check hash table.
4618          */
4619         write_lock_bh(&addrconf_hash_lock);
4620         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4621                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4622                         struct inet6_ifaddr *bifa;
4623
4624                         bifa = ifa;
4625                         ifa = ifa->lst_next;
4626                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4627                         /* Do not free it; something is wrong.
4628                            Now we can investigate it with debugger.
4629                          */
4630                 }
4631         }
4632         write_unlock_bh(&addrconf_hash_lock);
4633
4634         del_timer(&addr_chk_timer);
4635         rtnl_unlock();
4636 }